1)It is important to point out that the Gemara does not actually use the words "beginning" and "end", but rather the "Ikar" - i.e. the "principal" part of the Berachah - and the end part. This is because the beginning of a short Berachah contains the principal parts of it, i.e. the 2 names of Hash-m and the "King of the world", while the end contains no names of Hash-m but only the specific words for the type of food being consumed. Therefore there is a strong Sevara to say that since when he said the Ikar part of the Berachah he mistakenly thought that it was wine, it is not sufficient that right at the end he corrected himself.

2) Your question seems to be what the Gemara itself meant when it rejected the proof from Yotzer Or and Ma'ariv Aravim - i.e. that one can differentiate between a short Berachah and a long Berachhah. Rashi DH SHA'ANI explains that the difference is that with a long Berachah, since one makes a different Chasimah - of Baruch Yotzer ha'Me'oros - he is Yotzei because of the correct Chasimah. In contrast, with a short Berachah, since he said the names of Hash-m with the wrong drink in mind, he has no opportunity afterwards of a Chasimah with which to correct this mistake.

3) From Rashi DH ELA it would seem that it is not possible to change the conclusion of a short Berachah if the wrong conclusion was said the first time. One sees this from what Rashi writes that he started off the Berachah with the intention of concluding with the Berachah for wine, but before he reached the end he remembered that it was beer and concluded with the Berachah "Shehakol" for beer.

The Rashba points that one learns from Rashi that if he would only have remembered after he had already said "Borei Pri ha'Gefen" and then immediately remembered and said "Shehakol", this would certainly be no good. The Rashba writes that the reason is because since this is a short Berachah when he says Borei Pri ha'Gefen this is considered that the Berachah is finished so if he then adds on Shehakol this is considered as an addition after the end of the Berachah and not as a Chasimah in its own right.

(However afterwards the Rashba cites the Geonim who disagree with Rashi and maintain that one can change the conclusion of a short Berachah after saying the wrong conclusion).

KOL TUV

D. Bloom

I think the sugya on 12a is probably the main one for Kavanah on food Berachos but there are other sources.

(1) The Mishnah 40a states that if one mistakenly said ha'Adamah on the fruit of the tree the Berachah is valid. However Rabeinu Yonah (28b in Rif pages DH AVAL) writes that if someone made ha'Adamah and afterwards a different species, whose Berachah is ha'Eitz, was brought in front of him, he must make a new ha'Eitz and is not covered by the original ha'Adamah. R. Yonah writes that the aforementioned Mishnah, that b'Di'eved ha'Adamah covers ha'Eitz, only applies if at the time one had Kavanah to cover the fruit of the tree, but if when he made ha'Adamah he did not have Kavanah for the fruit, this is not valid.

There is an interesting practical example of these concepts. Mishnah Berurah 202:76 writes that one should make "Shehakol" on sugar but b'Di'eved if one said ha'Adamah one is "Yotzei". Similarly "ve'Zos ha'Berachah" by R. Alexander Mandelbaum of Har Nof, Jerusalem, p.103, writes that some contemporary authorities maintain that even though l'Chatchila one should say Shehakol on chocolate, nevertheless b'Di'eved if one said ha'Eitz or ha'Adamah one is covered. Therefore one should first make Shehakol on the chocolate or sugar and only afterwards make ha'Eitz on an apple because if one does it the other way round, it may be that the ha'Eitz on the apple covered the chocolate or sugar and one can no longer make a Shehakol on them.

(2) The Gemara (end 53b) states that Shmuel asked Rav whether one should say Amen after the Berachah made by young children (learning the Berachah from their teacher - Rashi) and Rav answered that one answers Amen after every Berachah except from that made by children because they are only making the Berachah in order to learn it. Rashi DH Ho'il explains that the children do not have Kavanah to make a Berachah. One sees from this that a Berachah made with a different intention in mind is not considered a Berachah.

Shulchan Aruch OC 215:2 adds that if the person making the Berachah is a heretic one also does not answer Amen because he does not intend to make the Berachah for the sake of Heaven, but rather for his idol-worship.

(3) There is another interesting Rabeinu Yonah (6a in Rif pages DH AMNAM ) which is actually in our Sugya on Gemara 12a. There is a dispute in the Gemara below 13a and Eruvin end 95b whether Mitzvos require Kavanah. R. Yonah writes that this dispute only applies where there is an action involved - for instance shaking the Lulav on Sukos - where there is an opinion that if one lifted up the Lulav this constitutes a Mitzvah even if there was no intention for the Mitzvah. In contrast, for a Mitzvah which depends only on speaking, everyone agrees that one requires Kavanah, because the crucial part of what one says is that one thought about what one said. Therefore if one did not think about the Berachah one said, this means that one did absolutely nothing, because there is neither an action, nor any Kavanah in one's heart.

KOL TUV

D. Bloom