Deja vu multiplex

To see how Déjà vu handles multiple translations is unfortunately a Déjà vu experience. I tried this for my almost finished Studio project; export the memory and retranslate/proofread in Déjà vu. There is of course some loss of hits, as segmentation differs, but that was in the beginning compensated by the very nice Auto assembly/Lexicon function that picked up parts of the segments from memory and in many cases recreated the missing TU:s very nicely. But then I run into a serious problem and according to list mates on déjà-yahoo that is the way it is supposed to work and thus unsolvable.
The problem is that when I make an edit during proof to a 100 % match, Déjà vu for some reason produces a multiple translation in the memory! I think it was acceptable to correct 10 % of the sentences during proof, but to manually remove a multiple translation for every edit change in a 40 000 word proof would simply not be possible. Neither can I accept to have a database where a lot of the units are incorrect.

After making comprehensive tests with both Studio and DVX, I do not understand the logic behind the solutions that CAT programmers make. They are in many cases contra productive and an obstacle for the translator, instead of being a support in the translation process.