If you have to pitch shift plus/minus by more than ~4 semitones (M3), then you are probably better off just re-recording the vocal track in the different key. You'll never be able to reproduce the harmonic spectrum of the vocalist correctly after pitch shifting by large amounts. To do that you would need a model of the performers vocal track including how its resonances/formants vary with frequency and amplitude. Besides a harmonic exciter, you could also try using a match EQ where the spectrum to match is the original un-pitch-shifted vocal track. So keep a copy of the original untranspose track, and use that as the spectrum to match while processing the transposed track. Then after the match EQ, maybe try the harmonic exciter. I'm not sure what order would work best, but theoretically it would seem to make sense to do the match EQ first, and then the harmonic exciter if needed.