OK, I have checked through things again. I am happy with the results for the 8A, but I may have found an issue with the Ultralite results.
For those who are unfamiliar with this type of interface they have become extremely capable and sophisticated. They have mixers, effects like reverb and filtering all performed in on board DSP. They are essentially mini recording desks / studios. You can of course bypass all of this.
Below is the Motu interface. The input settings can control input gain from -96dB to +22dB. Output settings can control output volume over a range 0g -127dB to 0dB.
You have extensive signal routing, essentially you can send any input to any output via any effect or mixing along the way.
Mixing desk
It all might seem terribly complex but once you get used to it its fine.
Now back to the issue, I may have not had the settings set back to the simple "measure" parameters which is just in/out with no other processing. A mixer fader may have reduced the output level. However hopefully all this will do is move the level of the lowest IMD. It would explain the different harmonic distortion levels from the original measurement.
Second issue. In the original measurement I discussed the spray of harmonics at high level which was disappointing. It looked like amp clipping, which is exactly what it is. With both the input and output gain settings at 0dB both input and output clip.
With a fixed output DAC this would be horrendous, but thats not what this is. We can legitimately change the gain of the input and output amps turn the volume down!
It appears that the output needs to be set to -1dB and input set to -2dB to prevent clipping with a looped back 0dB input signal. See below. Basically Motu have screwed up the gain settings on the amps. The output set to 0dB should not clip. To be clear this is an output voltage of 5.980 V RMS (balanced) and half that single ended. Way above the normal domestic voltage of 2 V rms. This is probably not to be considered any issue for the input as when recording you would always monitor the input level on a VU meter anyway.
I dont think its "cheating" to run the output and input at slightly reduced levels as being variable is part of their normal function. Its just annoying that something simple like this could cause problems if you dont know about it.
Note that this is the old version of the Ultralite mk4 with AKM DAC chips. Its going to be interesting to see if the new version with ESS Sabre dacs has the same issue. The 8A doesnt appear to suffer this issue.