1. The measurement shows the amount of EMF of the headphone driver (speaker), existing at this point of measurement. The differences in measurements are purely and only caused by the output impedance of the used amplifiers.

2. The amplifiers do not distort at all. The 'distortion' is visible because all the amplifiers don't have have zero ohms, but 10 Ohms, or - in case of the ADI-2 Pro/DAC - around 0.1 Ohms output impedance/resistance. Yes, this is already enough to get strange looking graphs. One could easily measure inside the box, directly at the amp's output, before the internal protection resistor of 0.1 Ohms and the relais (yes, even that one has 0.x Ohms and makes a difference) to find that the amp itself still delivers a fully clean signal.

3. Talking about the alleged THD effect: From a developer point of view it is disappointing that the author of the TI paper published such a misleading graph in a serious technical paper. As it is that so many people in said thread did not take the time to read the paper to understand it. The main point of the paper is how to make standard op amps stable for capacitive loads, and protect them from oscillating at very high frequencies. This, of course, is no issue if you use circuits that have been designed already not to act like this, and are developed specifically for operation with phones.

Just measure the signal at the phones driver itself, after the headphone cable, and you will notice that the 'distortion' is magnitudes higher again, and the output impedance of the amp now does not make any difference anymore. Especially the difference between an ADI with 0 Ohms and 0.1 Ohms gives zero difference.

This kind of measuring, THD at the amp output with a driver connected, is misleading, and does not give any useful information about the stability, behaviour or sound of an amp. And this is the sole reason this kind of measurement is found nowhere - it does not serve any purpose.

The reason to design amps with lower output impedance is mainly to avoid changes in sound with some phones/in-ears. Therefore measuring the frequency response at that connection point makes sense (just not THD).

Technically this all comes down to the damping factor, which also has been abused for misleading technical specs and alleged sound behaviour. One should read the JBL paper about 'The Damping Factor Debate' (Augspurger, already published 1967) explaining why damping factors above 10 do not make any difference, and why many companies calculate the value in a wrong way. There also exists a newer paper from Solderdude (Resistance, Impedance and other issues) who tries to explain the same with a view on the currents flowing.

Finally the measureable EMF, even directly at the phones driver, does not correlate with the acoustic output of the phones driver. That would be the only way to measure distortion differences of amps, but it always fails because the driver exibits values magnitudes higher than the amp.

So if you want to measure distortion use resistive loads, and -  these days - do it with 32 Ohms to stress the amp. If you want to measure the real output impedance use resistive loads as well. Using a headphone for these is just wrong.

Thank you for your replies and explanations. Highly appreciated.