If you're using a UST, the result will only be as smooth as the OTO of the CV voicebank. It's more practical to use the CV voicebank directly so that you aren't passing the samples through a resampler twice.
Ideally, you can record a new VCV voicebank from scratch. CVVC Japanese is also an option, which requires much less recording and otoing than VCV, but it's still smooth and easy to use.
If you recorded the original CV voicebank a long time ago and and can't do the same voice today, you can train a voice changing model on the old samples. That way, you can record a VCV reclist and convert the audio to sound like your old voice. Only do this with your own voicebanks or voicebanks that give you permission.
If you can't record anything at all and want to just use the old voicebank samples, I recommend editing together the raw samples. That way you can do more careful detailed editing, and you can ensure that the raw audio samples are only going through UTAU once when you use the final VCV bank.