If there's one thing I learned from Deci's Japanese voicebanks, it's that power isn't everything. When you find a song she works in, she does it well, but she's also a lot more limited than one would think. That's why I would say that for versatility, a soft voice is better than a powerful one (and multiexpression being best case scenario).
This is also one of the main issues I'm addressing with her English voicebank, the other being not rushing it like I did Deci Mal 64 JPN. As for sound quality, the voicebank will still have noise, there's really nothing more I can do about that with being an amateur at sound design, recording, etc. Although, at least, like recent previous takes on the English voicebank, I'm no longer recording with the microphone feet away from me (which yes, was how I recorded Deci's previous voicebanks) so the recordings won't have excessive reverb.