Minor tweaks again. Time for a status update:
After all these tests, try after try, there are (as always, but now, a lot more verified), two possibilities:
- We *can't* get realistic reflectance spectra from RGB Triplets without some other information.
- We can get realistic reflectance spectra.

The latter is what I hope :-) and I discovered I'm not the only one that worked on this: there are at least other two (old) papers about calculating a Reflectance Spectrum from RGB values. This is conforting, for two reasons: I'm not the only one that think it's possible; the guys of the papers got the same problems I'm facing, so I'm on the right path :-) !