I think the 'phonetic borderland' hypothesis is surely possible. But I also think that trying to decode the VMS by multi-lingual phonetic assonances is hopeless, there are so many ways to do this that one is bound to find 'solutions' which are due to pure chance. And it's especially hopeless when starting from an arbitrary assignment of phonemes to glyphs, such as EVA.
By the way, in your phonetic reconstruction you also made the additional hypothesis that EVA 'q' = EVA 'k' (= EVA 'h').
By the way, in your phonetic reconstruction you also made the additional hypothesis that EVA 'q' = EVA 'k' (= EVA 'h').