**) As of now I have no explanation for this effect. Below a comparison of an SMSL SD1955+ and an RME ADI-2 Pro FS R with the exact same settings and conditions for the measurement, using the analog bandpass. Test frequency was 725Hz to allow maximum clearance to both mains hum/buzz and the idle tones of the SMSL (coming from the charge-pump generating the negative opamp supply rail).
Note that at least for the RME the plot is noise-limited, partly induced by the setup (it gets a bit better with the steeper digital bandpass). For the RME I've found, in a more elaborate, not noise-limited measurement, that it does what any D/S-DAC is supposed to do: linearity 100% nailed down to bit 24!