not an unpopular onion i just had this come to mind when i glanced through-- but its interesting i never really see ppl think about or even theorize on why an AI voice would sound so different from its previous non-AI voice, because it is kind of typical for no-AI voices to basically reflect like.. The key sound of that person with no influence on how they may sound normally- talking or singing in an outside context that isnt short samples? thats basically why. i feel like the most blatant comparison was Dex compared to his provider. theyre even featured in songs together and theres many parts where i can hear "dex" in Lupin, where dex as constantly reflects that unique tone through the song. at first they sound different until you pick it up in Lupin's voice.
with flower- i dont think Ci Flower was actually trying to be that different from v flower, have you heard her x voice samples? the spoken recordings? i remember getting those and thinking "wow that doesnt sound like flower at all" and they were the first example of what her VP sounds like outside of the vocaloid bank. Her voice samples sound WAY more akin to her Talk bank and later on Ci flower. this shouldnt exactly be a shock to hear because its been showcased before. Vocaloid flower took that unique tone that appears sometimes, and made it constant, because every individual sample made for a non-AI is way more likely to do so.
it happens with my own UTAU and im sure it happens with everyone else, as much as the bank still sounds like you. that is like, the exact reason any of these banks would sound more and less like each other- even a voice as unique as Teto managing to get the right sound, its still different and i feel the places where it is different are for that reason. the samples for AI voicebanks will never be as "exact" as the non_AI ones (they could potentially, with different or more data of some kind).