That’s normal. You can also experiment with shifting the voice up or down an octave, Auto Tuning, and then putting it back to the original register.
But Auto Tuning applies observed mannerisms to the vocal. They are - by their nature - human imperfections. For now, I think they’re best seen as a way to “humanize” the performance, and not a way to get a more “expressive” performance.
I suspect Auto Tuning is still at a fairly early development stage. With more songs or focus on specific styles, it’ll be better.
In the meantime, you’ll want to go back and manually add expression using the standard tuning tools.
And don’t forget to turn Vibrato Depth back on when you’re using the regular SynthV vibrato!
I should also mention that - in my limited experience - as far as “expression” is concerned, Tension, Gender and Loudness tend to be more important than pitch.
For example, if you want to give the impression of vocal effort, increasing Tension and Loudness are helpful. Singers also open their mouths wider when singing louder, which changes the shape of the vocal tract. This increases the frequency of the first formant - something you can simulate with Gender parameter.
None of these parameters are currently handled by Auto Tuning, and frankly, are a bit harder to measure and train a model on, so I’m not sure that I’d expect Auto Tuning to incorporate in the near future.
All of which is to say - Auto Tuning will get you part of the way somewhere - but to get a good performance, there’s still a lot of manual tuning to be done.