Del Vento Interesting, I never thought about that, that's really interesting.
I think it is due to the wavelength of the fundamental frequency of each key. A tone one octave higher has half the wavelength, and there is simply less "space" for vibrations than one octave lower. Therefore tones with higher fundamental frequency have less complicated sound than those with lower fundamental frequency.
EDIT: I can explain it better this way: when the sample is transposed one octave up, it is played in half the time. Therefore, a lot of very high-frequency material is added, beyond the range of the original sample. Or, to put it differently: the entire power spectrum is shifted up towards higher frequencies, thereby taking on a completely different sound. Likewise, when shifting one octave down, the power spectrum is shifted towards lower frequencies, so a lot of high frequency material present in the original sample is no longer present.
Del Vento Try two octaves for harpsichord? I wonder if this technique (used by just a few notes) can be used instead for "voicing" a piano slightly more bright or dark, without making it another instrument!
A problem in going one octave further is that you loose one further octave towards the bass end, but the idea is intriguing. The "voicing" idea is interesting; my gut feeling is that you are right about it.