0dBu (=0.77Vrms) is a weird level to measure at and complicates comparisons to other measurements.
The performance at this level (98dB SINAD) is pretty good, but it's unclear what part of that is the signal generator, what part the ADC, and what part the actual headphone Amp.
Yes, absolutely wrong! Maybe I will put more light on my hardware setup:
I used Motu M4 with SE output and BAL input.
The measured amplifier input signal is about 1Vrms, which is corresponding with my listening setup (voltage from pre-amplifier).
I listening on Shure SRH-1840, so I used 64 ohms in REW setup.
Amplifier has selectable gain which is now x2.
I finally calibrated REW input and output voltage and it looks more predictable now.
I mean maximal gain (~500mW, ~17dBu) correspond to input level provided by Motu (+18dBu) and full scale of M4 input meter.
Below measurements for 10mW (0,4dBu), 50mW (7,30dBu), 150mW (12,08dBu), 300mW (15,08dBu) and 500mW (17,26dBu):