Are you sure? He quite clearly has it marked that he took measurements, including SINAD, with 300 ohm and 33 ohm loads, with those impedances matching the values at the very bottom of his dashboard screenshots.
Fair enough, Yes I just looked at the first one, but really the one that matter is at 33 ohms for this, and it’s quite low. I find your threshold very low for modern use cases. you realize that in real life this SINAD is only true if there is no digital attenuation in the chain. and this 75 dB range would be from a 0 dBFS peak, where your RMS levels of most recordings are about 20 dBs below that. In real life, with non ideal gain structure, a 75 dB SINAD can mean quite audible noise and distortion in many use cases. But again that's not all, other stuff matters, the thing is we have all these measurments for the linear behaviour, more often than not at 1k, but it's rare to see measurments toward the limits, and rightfully so, Amir could break most of the devices he received. We tend to think that our device always operate in their optimal zone, but you'd be surprised how many peaks are clipped or compressed with micro-powered device with normal headphones. You'd be surprised to see some frequency response, or IMD graphs if they where taken at the limit of available power, or or with dynamic loads. We like to see margins, headroom, because they are used more often than we think. People sometimes refer to dynamic range of records of tape to assess that all those extra bits are pointless, but fail to recognize that in the analog days, there was no 105dB/mW IEMs, the calculations where simple, a room, speakers, environmental noise floor, dB SPL at listener position, easy to optimize gain structure. Now the use cases and type of transducers are all over the place, often great amount of digital attenuation, gain structure all over the place as well, and with portable device like this, heavily compromised power availability. So looking at a poor SINAD value and conclude that it's transparent anyway is bogus most of the time, it's only transparent in very specific optimal cases that rarely occurs