JA always uses the same measurement methods for all DACs. I can see the value in that.
However those measurement methods don’t work well in the presence of noise. For example: resolution is the smallest change in input that has the correct (or, some say, any) change in the output. The resolution of the DS and DS Mk II are both much higher than JA quotes. In the original DS Mk I he even complained about some non-linearities down around -120dB and then could see the differences with later software releases. If the resolution were indeed only 18 bits, then how was he seeing measurable differences down near 20 bits?
Another example is that the audible noise on the MK II is clearly better than the MK I, not worse. The differences being caused by the bandwidth on the AP used to measure the noise.