Where V-Vocal falls down (and Melodyne shines) is in the pitch detection. V-V divides the clip into 50ms segments and determines the fundamental pitch for each one. If it miscalculates some of them you'll hear artifacts as bogus frequencies from adjacent segments beat together. The result is a particularly offensive metallic phasiness.
Even if you simply bounce the clip without making any edits, the artifacts will still be there, because the damage was done as soon as you created the V-Vocal clip.
To get the best results from V-Vocal - and good results really are possible - you have to help it out. That means giving it clean, loud tracks with minimal background noise and headphone bleed. It also means segregating out portions of the track (split and trim) that need correction the most, while
not including the good-enough portions in the V-Vocal clip that don't really need editing. In short, the higher the quality of your tracks the better job V-V will do.
As Craig points out above, you don't want to include non-pitched sections for correction because V-V will never figure out what their fundamental pitch is and artifacts will surely ensue. And it may not be enough to just not edit them: ideally they should be excluded from the V-Vocal clip altogether.
The worst mistake you can make is automatically creating V-Vocal clips for every vocal clip and determining which notes need correction by looking at the V-V display. Don't do that. Ever. Always listen to the clips first and decide whether they
sound out-of-tune or not.
Do not assume that straightening out the pitch will make them sound better! So yeah, compared to other pitch-correction tools, V-Vocal can be a pain. But there are many users for whom better alternatives are not an option. For those folks, it's good to know that V-V can in fact do the job most of the time, with a little extra effort.