First, some background:
The idea that each traditional berakhah and tefiyllah that can be found within the prayerbook has a specific number of words - even a prescribed number of letters within each word - and that each one is thereby meant to express a mystical meaning through gematriya (thus "proving" that it was transmitted perfectly from the time of the tannaiym or earlier) is an idea that stems from Eli`ezer b. Yehudah of Worms in his Sefer Roqeyah. So entrenched is he in this idea that he states that one who changes even one letter of the nosah [Ashkenaz] is not accepted in Shamayim when he prays. This idea has now been proven by extant texts to be patently incorrect, as there are many ancient usages that all share a common core, the codifiers of which apparently paid no attention to either word count or gematriya - something already noted by the Beyth Yosef (cf. Tur, OH:113-122 - comments based on the Abudaraham).
As for the view of the Rambam, one must remember to always carefully review the the construction of the text, as his choice of phrasing are often overlooked. His methodology is to both preserve the wording and terminology of the original Talmudic source(s) which he is codifying, as well as be exact regarding the nature and definition of the halakhah as he understands it. Also, the Mishneh Torah was intended by the Rambam to be taken in aggregate, using logic to sort out the path of the halakhah from all relative statements made regarding it throughout. Lastly, it is a well-known interpretive principle of Hazal that when they (and subsequently the Rambam in recording their words) make "all" or "never" statements, there is nearly "always" an exceptional case.
Noting these things, let's re-examine the halakhah cited above.
The Rambam in Hilkhoth Berakhoth 1:5-6 states the following:
 wa-nosah kol ha-berakhoth 3ezra uveyth dino tiqnum wa-eyn ra'uy le-shanothan wa-lo le-hosiyf 3al ahath mehen wa-lo lighroa3 mimmenah wa-khol ha-meshaneh mi-matbea3 she-tav3u hakhamiym ba-berakhoth eyno ela to3eh wa-khol berakhah she-eyn bah hazkarath hashem umalkhuth eynah berakhah ela im ken haythah semukhah la-havertah:
"And the wording of all the berakhoth was formulated by Ezra and his court, and it is not proper to change them - neither to add to them nor to subtract from them, and everyone who alters the 'coin' minted by the Sages is simply mistaken. And every berakhah that does not have the explicit mention of the Divine Name and the Divine Sovereignty is not a berakhah at all, unless it is semukhah le-havertah ("next to her neighbor" - i.e. directly follows a berakhah that does include the mention of the Divine Name and Sovereignty, which are considered to fulfill this requirement for the following berakhah as well)."
 kol ha-berakhoth kullan ne'emariyn be-khol lashon wa-hu she-yomar ke-3inyan she-tiqenu hakhamiym wa-im shinah eth ha-matbea3 ho'iyl wa-hizkiyr azkarah umalkhuth wa-3inyan ha-berakhah wa-afilu bilshon hol yassa:
"All of the berakhoth may be said in any language, as long as one says them in keeping with the subject determined for them by the Sages, and if one altered the 'coin,' as long as he made explicit mention of the Divine Name, the Divine Sovereignty, and the subject of the berakhah - even in a language other than Hebrew - he fulfills his obligation."
In a similar passage, the Rambam also says (Hilkhoth Qiryath Shema3 1:8):
 berakhoth ellu 3im she'ar kol ha-berakhoth ha-3arukhoth be-fiy kol yisra'el 3ezra uveyth diyno tiqenum wa-eyn adham rasha'iy lifhoth mehen wa-lo le-hosiyf 3aleyhen: maqom she-hithqiynu lahtom be-varukh eyno rasha'iy shello lahtom umaqom she-hithqiynu shello lahtom eyno rasha'iy lahtom: maqom she-hithqiynu shello liftoah be-varukh eyno rasha'iy liftoah umaqom she-hithqiynu liftoah be-varukh eyno rasha'iy shello liftoah: kelalo shela-davar kol ha-meshaneh mi-matbea3 she-tav3u hakhamiym ba-berakhoth harey zeh to3eh wa-hozer umevarekh ka-matbea3:
"These berakhoth - along with all the other berakhoth that are arranged in the mouth of every Jew - were formulated by Ezra and his court and it is not proper for anyone to shorten them or to add to them. In a place [within the overall corpus of berakhoth] where they determined that a berakhah should have a hathimah using 'Barukh..." it is not proper to leave it out and where they determined that a berakhah should not have such a hathimah it is not proper to add one. In a place where they determined that a berakhah should not have a pethihah beginning with 'Barukh...' it is not proper to add one, and where they determined that a berakhah should have such a pethihah, it is not proper to leave it out. The general principle of the matter is this: whoever alters the 'coin' minted by the Sages with regard to berakhoth, behold such a one is in error and he must go back and pronounce the blessing in accordance with the 'coin.'"
At first glance, these two statements by the Rambam may appear to contradict each other, but keeping in mind the style and methodology of the Rambam we mentioned before let's parse out their meaning.
- The 'coin' that the Rambam mentions is the word matbea3 (מטבע) and it is not his own invention. It is employed by the Tosefta (t.Berakhoth 4:5) and is subsequently found in both the Talmudh Yerushalmiy (y.Berakhoth 1:5; 5:2; 6:1-2; 9:1) and the Bavliy (b.Berakhoth 40b) as a baraytha. It is a reference to the formulation of the berakhoth; like each coin that is minted might have some internal differences, yet they all share the same basic structure and elements. So, matbea3 may be understand as the "essential nature/structure" of a berakhah. There are several things included in the matbea3, which I will mention below.
- The word nosah ("formulation"/"phrasing") used here is a general term used for the overall concept and collective elements that make up berakhoth which were decided upon by Ezra and his court. It should not be understood according to common usage today, as a reference to the exact wording of each berakhah.
- The Rambam does mean to imply that since the time of Ezra there have been no changes or additions to the body of berakhoth. This may be seen from the fact that the Rambam himself talks about Tannaiym drafting berakhoth and elements of tefiyllah long the time of Ezra. However, as we shall see, these later hakhamiym possessed proper legal authority to do so and they also were constrained to follow the pattern of formulation for berakhoth that was initially legislated and determined by Ezra and the Anshey Kenesseth HaGedholah.
From the Rambam, we can extrapolate several things:
1 The "formulation" (nosah) of berakhoth includes three elements:
- (a) The 'coin' (matbea3)
- (b) The Divine Name ("...Adhonoy elohenu...", referred to by Rambam as the azkarah, which according to Rav Yosef Qafih z"l and others indicates that the Rambam rejects the use of kinuyyim for HaShem in other languages - i.e. he requires that the Hebrew Names be used even when the rest of the berakhah is in a language other than Hebrew)
- (c) The Divine Sovereignty (referred to as the malkhuth - i.e. "...melekh ha-3olam...")
 The 'coin' or matbea3 itself contains two elements:
- (a) The 3inyan or subject matter of the berakhah as determined by the hakhamiym.
- (b) The pethihah or hathimah (or both).
From here we can see that the view of the Rambam - which is also commensurate with the textual evidence - regarding the altering of berakhoth is that while the mention of the Divine Name, the Divine Sovereignty, and the presence of a pethihah or hathimah (or both), may never change and still have the mevarekh be yosse yedhey hovah, the 3inyan - although it should not be done on purpose - may mistakenly be changed and the mevarekh is still yosse yedhey hovah. This is clear from a careful reading of the texts. The Radvaz notes that when the Rambam says, eyno ela to3eh he means that such a one "is merely mistaken (and does not have to repeat the berakhah)" whereas when he says harey zeh to3eh he means that it is a complete error and thus one must repeat the berakhah according to its proper form.
This also fits with the Tosefta (t.Berakhoth 4:5) on which the Rambam is basing himself which says:
ר' יוסי אומר כל המשנה ממטבע שטבעו חכמים בברכות לא יצא רבי יהודה אומר כל שנשתנה מברייתו [ושנה ברכתו] יצא
"Rabbi Yosiy says, 'Anyone who alters the 'coin' which was minted by the hakhamiym with regard to berakhoth does not fulfill his obligation.' Rabbi Yehudhah says, 'Anyone who changes the nature (i.e. the subject), [and in effect changes the intent of his berakhah] still fulfills his obligation.'"
These types of things are noted by Rav Sa3adyah Gaon in his prayerbook. There are a few places where he notes that he is aware of certain groups of people making additions to the wording of certain berakhoth; in some instances he says that although it is not in accordance with the strict requirements of the halakhah, it nevertheless does not damage them in fulfilling their obligation, and in other instances he notes that such changes - even minor ones - serve to change the original meaning of the berakhah. One such instance is on page 21-22. The Gaon's language there is instructive of what has been said here.
Further proof of this understanding of the Rambam (cf. Hilkhoth Berakhoth) can be seen from his explanation of the four blessings of birkath ha-mazon and what needs to be mentioned in order to fulfill one's obligation. Also, when explaining the tashlumiyn (the 'make up' for a missed prayer - cf. Hilkhoth Tefiyllah), the Rambam advises to make a slight addition to one of the middle berakhoth of the shemoneh 3esrey (called "hidhush davar") in order to distinguish it from one's tefiyllath hovah. And there are other such examples to be found in the Mishneh Torah that are similar to this.
The rishoniym and aharoniym who altered berakhoth, usually did so based on variants mentioned in the literature of Hazal or for kabbalistic reasons. The Arizal and other kabbalists felt that they had secret doctrines and insights that were worthy of the halakhah as transmitted in the Gemara being overridden. They had their reasons, but as the Rambam said: eynam ela to3iym. There were those, such as the Rashba, who felt that the matbea3 only included whether or not a berakhah had a pethihah or hathimah, and this also contributed to later innovation.
As for the specific examples you mentioned:
The fourth berakhah of birkath ha-mazon was added by hakhamiym with proper legal authority to do so.
"she-`asani kiressono" and "ha-nothen laya'ef koah" both come from Ashkenaz and are very late additions. Even the Shulhan 3Arukh says not to say them because they have no basis in the Talmudh. The rishoniym - and especially the Rambam - rejected many other berakhoth that were created by the geonim for the same reason.
I think you mean the berakhah for Yom Tov candles, which is not cited by the Rambam - which is why many Temaniym do not make a berakhah when lighting them.
Hope this has helped.