There’s nothing special about 100 or 99. They each map to a linear volume with lots of ones and zeros. Each different upsampling filter has many coefficients with many ones and zeros and further, depending on the sample rate, there’s a volume correction based on the upsampling factor, e.g. tho samples at 44.1k are scaled by 1.000976174753312 samples at 48k are multiplied by 1.089497877282516 (which comes from the squareroot of 6.143632880538761e-9 * 147 or 160 and binary shift.)… So there’s no input volume that has mostly ones or mostly zeros. Then the sigma delta modulator is chaotic and hence every possible DC level (even being muted) has a time varying pattern of ones and zeros.
Don’t get me wrong, I can believe that some people might hear a very subtle difference between different volume levels (even if corrected for with an analog volume control downstream.) It’s just that at that level of subtlety every different volume level and sample rate would be different and none of them would be special, e.g. more or less technically faithful to the source.
I’ve got to believe that the differences from different volume levels of analog distortion in your preamps, amps, and, say, the transformer in the DS are more significant.