Amir, as I see your setup has too high common-mode noise or AP analyzer has a problem with its CMRR. I took a look at the result here
https://www.audiosciencereview.com/forum/index.php?threads/hidizs-s9-review-headphone-adapter.21152/
The balanced is a lot less noisy vs unbalanced, that's why my balanced 9038S looks so clean vs unbalanced 9038D. You can check that by shorting one of 9038D outputs(of course better if no signal on that channel), I bet you'll find the same dirt on the shorted output as well as on the working one. To avoid that noise you can try to use Android phone in airplane mode and battery-powered with a free Hiby hires player.
My measurement was performed in different ways and always I see the same result +/- a little bit. Regarding Dynamic Range, APx555 has 8x times more noisy inputs vs my custom preamp, hence it is fine if your result up to 1db worse(i.e. 124db(A)) but not 6db of course. The same about THD+N my SYS2522 shows me -117db, SYS2522+external ultra-high performance notch filter(the current 3rd version of that filter has -126db limit for THD+N) 9038D has THD+N -118.3db(-.2-.5dbfs, no-load condition, 20-22000Hz BW 0.00012%). Also, my new project Cosmos ADC(SNR -127db(A) in the mono mode) also shows me THD+N -118db. All my analog processing instruments(120nVrms(A) preamp and notch filter)are well-matched with their preliminary simulation and school physics too(I did test thermal noise of resistors to confirm the noise level resolution).
PS: I want to show how Cosmos ADC sees the signal of 9038D 12kHz@48kHz. No AVG at all. The FFT shows jitter of Cosmos+9038D.
View attachment 119260