On second look, one can calculate the noise floor that Stereophile likely measured, based on this sentence:
"The actual distortion doesn't rise above the
noise floor until the output reaches 15V, when it is just 0.0005%, but rises rapidly above that level."
15 V x 0.0005/100 = 0.075 mV which is -106dB, not far off the Bryston spec.
I am measuring 100 times that (7.5 mV) and a DC offset of -13 mV (R) and +37 mV (L) on the RCA outputs. The XLR outputs have no DC.