Threshold picks a level, above which, signals stay the same, or get quieter(depending on the ratio setting), and below which, signals are unaffected. Period.
Signals below the threshold level stay the same until you increase the makeup gain.
When you compress, you can knock down the peaks using threshold and ratio controls, and leave the quieter signals alone, OR, you can quiet the peaks, then turn EVERYTHING left UP with makeup gain.
In short, the only thing you can do to vocal artifacts like sibilance and breath/lip noises, with a compressor, is keep them constant, or turn them UP.
After you wrap your head round that, you have MANY options for reducing vocal artifacts, including but not limited to, noise gate, expander, EQ, multi-band compression (not for beginners), masking, selective doubling, cut and paste editing, fader riding, automation, etc.