Thank you for a very interesting insight into the problem of expressiveness (or lack thereof) of samplers. The approach you suggest is IMHO definitely the way to go.
A question, though, concerning phase alignment of p to f samples for a correct crossfade. I don\'t think this can be achieved in the time domain, due to the intrinsic nonalignment of the harmonics of different samples, so it is not simply a question of jitter of a particular sampler. Even morphing in the frequency domain requires some form of stretching of the main peaks in order to align them, for realistic results.