@amirm, it is a bit difficult to tell from that third chart due to the plots overlapping, at least with my eyesight, but it appears that all three modules had the same issue with about a 3-5db discrepancy in distortion level between the two channels. Is this correct? If so, I'm not an engineer, but wouldn't that indicate that the issue is in the modules themselves, not the implementation? The $1,000+ 2 channel March Audio P502 also showed the same behavior, albeit without the hump, although it actually had 3-4db worse SINAD.
EDIT: Actually the March did have a hump, especially at 8 ohm.