The gemmorah tells us that although one made Havdalah in Maariv one is still obligated to repeat Havdalah over wine.
A few lines later in the Gemorrah quotes Rav Acha as saying that it is better to make Havdalah in Davening than over wine, but if one does BOTH he will get blessing on his head. The gemmorah questions this and says it seems from Rav Acha's words that one alone will suffice, if that is the case then why should he be blessed for saying Havdalah twice it should be an unneeded blessing and as such that should be a violation of LO SISSA! The gemorah retracts and says rather if he makes only ONE Havdalah it is better.
Now does Rav Acha argue with the gemora quoted above that said "although one made Havdalah in Maariv one is still obligated to repeat Havdalah over wine."? And even if he is, why does the gemmorah's problem of Lo sissa not apply to our accepted Psak which requires us to say Havdalah in davening and over wine? Also when the gemorrah changes Rav Acha's words all it says is that if one only made Havdalah once he should get many blessings. But that was just the end of Rav Acha's statement. How, according to this retraction would Rav Acha Start his statement? It seems his entire statement should be, one should say Havdalah in Davening and over wine, but if you only said it once than you should be blessed! Then why should he say Havdalah twice?
Tizku Lemitzvos
Yosey
Let's take your first question first. Rav Acha Aricha is quoting a Beraisa, so if anything, the question ought to be how the earlier Amora'im can argue with a Beraisa (not vice-versa). Perhaps it is because of the principle that any Beraisa that is not cited by R. Chiya and R. Oshaya is not authentic. I'm not sure why we do not always abide by it, but here it seems, we do, as we rule like the earlier Amora'im and recite Havdalah twice.
As for your question concerning 'Lo Sisa', we've already answered it. Because 'Lo Sisa' only applies where the Chachamim instituted only one B'rachah. Where they instituted two (such as here, according to the Amora'im), then one recites them both without qualms.
Incidentally, the reason that they instituted two B'rachos is because they really would have instituted Havdalah al ha'Kos to begin with, only the people could not afford it. So when the poeple became more affluent, and they did indeed introduce Havdalah al ha'Kos, they left the original Takanah (of inserting it in the Amidah) intact.
Rav Acha Aricha's Beraisa, on the other hand, seems to hold that when the Chachamim instituted Havdalah al ha'Kos, they did not institute it as an addition to Havdalah, nor even as a replacement. Instead they instituted it as an alternative(not wanting to negate the original Takanah), leaving us with the option of one or the other. More than that, having initially instituted Havdalah in the Amidah, they left that intact Lechatchilah (the Beraisa's first statement), only giving the alternative of Havdalah al ha'Kos as a Bedi'eved alternative (in other words, permitting and praising someone who recites Havdalah al ha'Kos, and not in the Amidah [the Beraisa's second statement]).
I do not understand your suggestion that the Beraisa's first statement is that one should say Havdalah in Davenning and over wine, when the Gemara has just ascertained that according to this opinion, saying it twice is a B'rachah she'Einah Tzerichah (because it accepts the first statement as it stands, changing only the second one).
Kol Tuv,
R. Eliezer Chrysler