The Benchmark DAC1 is not a "transparent device,"
This seems to go against the majority of conclusions and test measurements, as well as my own observations. FWIW - upsampling is used in many D to A devices. It would be difficult to explain Time Series Analysis or Signal Processing here (it is usually a 3rd of 4th year engineering) but I can assure you that upsampling will not change the signal but it can help improve S/N and remove reconstruction artifacts in the D to A, in a good implementation of course.