Not sure if this is your problem, but if you're writing the lyrics like
Code:
[か][え][る][の][う][た][が] or [ka][e][ru][no][u][ta][ga]
with just single diphones rather than
Code:
[- か][a え][e る][u の][o う][u た][a が]
the VCV voicebank wont work. It needs to have the preceeding vowel or dash to call the correct sample in the UST. The inclusion of the end of the previous vowel before the next lyric is basically the fundamental difference between CV and VCV.
If you don't want to do this manually, you can get a plugin like
Iroiro to automatically add the VCV prefixes to every note.