I have taken some new measurements with pre-out's also measured. The results are kind of interesting.
The first graph has 5 measurements, 3 towards the upper half that are in "loopback" -no microphone. I made sure these did not have the microphone calibration applied. This is essentially the correction Trinnov is doing. The colors are:
red: R972 in pure audio mode (no Trinnov)
magenta: R972 with position 1 selected, but EQ set to none
blue: R972 with position 1 selected and EQ set to flat
The remaining 2 are in-room measurements:
yellow: R972 with position 1 and EQ none
green: R972 with position 1 and EQ flat.
Note that the scale for SPL for pre-out and in-room may not be exactly the same.
The smoothing is set to 1/12 octave for the in-room measurements in this graph. The loopback (pre-out) measurements have no smoothing at all. Notice how smooth the loopback corrections are. Also notice how they get a little "fuzzy" after 10KHz.
That got me thinking, maybe this version of Trinnov calculates its corrections based on a certain smoothing factor. Perhaps due to a limited number of filters, or just because they know trying to correct in a higher resolution doesn't really make it sound any better. So I played with the smoothing on the in-room graphs to see if I could get one to correlate very closely to the pre-out corrections measured. What I arrived at was 1/3 octave smoothing. I'd say that's a pretty close correlation!
And here is a close-up of the "fuzzy" lines for the pre-outs, 15-20KHz. I doubt this is intended correction, but possibly some kind of noise or artifact of the correction. Note that pure audio is much cleaner than the no EQ and Flat EQ. Maybe the extra A/D and D/A conversion is part of it (does Pure Audio skip the A/D and D/A?)