I don't take blindfold tests very seriously. There are far too many variables involved. Somebody on another board was trying to convince me that there is no audible difference between 16bit and 24bit and cited some blindfold test as an example. He said the music was "orchestral" music, but when pushed further revealed that it was orchestral music using sample libraries. Obviously, recording real classical music in a real acoustic space like a hall is going to show off the qualities of 24bit (bigger dynamic range, better resolution for reverb tails, etc.) than a sampled, synthesized "orchestra", no matter how good the samples are.
For the piano blindfold, what sort of music was it? Playing a fast Beethoven piece with lots of runs and pouding of the keys is different than playing a slower piece with lots of sustain where strings are interacting with other strings causing sympathetic vibrations and such.