i'm not questioning your science purrin but i am sceptical that thd at the tiny, almost insignificant levels that your measurements record, could actually be heard to the extent that you believe. frequency response measurements on the other hand, do show the levels of treble, midrange and bass frequencies present in a headphone that are apparent when we listen to one. that doesn't mean that our hearing will necessarily align with what the measurements show (they should be regarded as indicative), but i do think that they are more relevant to the headphone listening experience.
Indeed the distortion levels at a few percent do seem miniscule, or at least the numbers do. However, as I said before and will repeat again: 1) Our ears are very sensitive to distortion, and 2) You have to understand what distortion actually is (I think you have some grasp, but not a good understanding).
To put distortion in frequency spectrum terms (which almost everyone should understand), here is another visualization on HP1000 vs. TH900 distortion. This time using two test tones at 22Hz and 66Hz near the reference levels used in the prior graphs. Ideally, a headphone should produce only these two tones. However, despite the "tiny, almost insignificant levels" (in your terms), why do we then see all this extraneous junk in the spectrum? That extra junk is distortion of a few percent or less. (I think what's throwing you off is how the percentages translate into actual levels.)
Download these two graphs and use a picture viewer to toggle between both to better see the differences. If you own or are familiar with these two headphones, you will know that the TH900 bass sounds so much cleaner than the HP1000's; and the HP1000's bass sounds much thicker, despite having a much less elevated bass frequency response than the TH900's "bloated" bass as the FR graphs would suggest! Frequency response does not explain everything.
While I do not contest your point that frequency response is one of the primary if not the prime indicator of how a headphone sounds like, other factors such as distortion, attack, and decay are what separates an LCD3 from say a SkullCandy Mix Master (which actually has a similar FR to the LCD3).
For additional reading, I would refer you to Zaph's site. He is actually places a very high priority on non-linear distortion, even more than I.
Edited by purrin - 7/21/12 at 5:51pm