I know this may be out of topic, and I apologize for that. However, since the 009 and 007 is being used in relative terms to describe the sound of the Abyss, this might be useful...
Below are the overlaid left channel FR measurements from the 009 and 007 using Tyll/Innerfidelity data (hopefully Tyll doesn't mind). The two are normalized so that the FR are matched at 1 kHz.
MH, while not the same set from the link you provided, the responses IMO seem match closely to those.
Seems to me the 007 has more bass than the 009 in this set of measurements. The 007 does have a steeper drop in the bass to sub-bass transition though.
Here is another view using similar scaling than Arnaud:
Now, regarding the Abyss distortion in the sub-bass region. According to Innerfidelity, it is ~1% at 20 Hz @ 90 dB/SPL, and below that throughout the sub-bass and bass region... and so is the 009. Pretty close IMO, so not sure what you mean. the Abyss does have more distortion in the mids and highs relative to the 009 though.
Innerfidelity linky to Abyss measurements:
http://www.innerfidelity.com/images/JPSLabsAbyssAB1266.pdf
Not sure if you realize that 1kHz is just an arbitrary point? Turn up the volume a bit on one and the graphs shift up. Basically they are the same and most likely within the tolerances of the measurement system. Pretty darn close, just like I hear it too. BTW, they are so close, I would hardly call one bass heavy and the other bass lite.
Have a look at the same pair of headphones measured by Tyll (before and after burn-in). Not sure how much of the variability in the bass to attribute to burn-in vs. headphone placement vs. measurement variability. There's not much variability here, but its close....just like the SR007 and SR009. When headphones measure this close, then I default to first hand subjective listening (which as stated they're pretty close indeed to my ears and on my rig).
http://www.innerfidelity.com/images/StaxSR009SZ91278.pdf
http://www.innerfidelity.com/images/StaxSR009SZ91278afterburnin.pdf
And with araud's measurements, I shall default to the author's comments and go with Tyll's data instead.