That is an excellent point.
Test Tones do NOT equal Actual Musical Notes, and we all know how complex music and sound can be.
Just because some people cannot discern a -20dB crosstalk, 1% THD, & +/-3dB deviation in Freq. Response of TEST TONES does not mean the same for actual music.
Actually, if you would have been reading the Journals and conference papers where the data has been presented, you would find that sensitive test tones are used because that gives the lowest threshold of detection.
Other signals, speech, music, etc, thresholds are much higher because of their complexity, masking, etc so the THD would be much higher,, like 1% with simple music and more with complex music, but also with the frequency bands involved. Bass is an example. Subs are rated to 10% distortion for a reason instead of 1% or .1%
Because no one can tell that 10% in subs.
Here, let me give you a link to Axiom's research:
http://www.axiomaudio.com/distortion.html#
In some cases 100% is not audible in bass
Now, with sensitive signals, this distortion is way down, well below 1%, .01 % perhaps.