The alternative theory about why the Blu2 CD sounds better is that the internal Blu2 CD player is a particularly low RF noise source . . . . .
To add to this, in my big Blu2 study, I did try wav files ripped from a CD vs lossless version of same file vs CD direct. The former two were the same, and the CD direct was the gold reference.
That gap reduced as I added ferrites to the USB cable, but I never got all the way to the gold reference of CD direct.
I no longer have my Blu2 (I have a HMS now), but my recent experiments with large buffer streaming end points point to a huge SQ lift via USB. I would be very curious to repeat my experiments from a year and half ago, but use a large buffer end point to compare with CD direct. I suspect the gap would be materially reduced.
Also on my list is to compare large buffers+USB vs optical, on the theory that an optical CD transport connected to HMS should be identical to Blu (operative word is "should"...I have not done this experiment yet) The optical input on the HMS is a wonderful feature and experimental lever, so I hope to get to this soon.
For me, all this is hinting at a combination of RF factors (as Nick suggests),
and some other (perhaps) timing or IRQ related factors on the end point.
My working hypothesis right now is to eliminate and isolate as much RF and electrical noise as possible (benefits are very clear here), and (more speculatively) isolate and optimize and get as close to real time (more precisely, zero variability latency) the digital input to HMS.
Every tweak I'm making that falls into these areas is rewarding me with a noticeable SQ lift.
Lots to explore and learn.