Interestingly, direct subtraction in Adobe Audition makes better job in this test than the Deltawave. Deltawave delta file is much more audible than the directly subtracted file, in this test. This is interesting and I would like to know what makes the algorithm to create bigger difference.
Can you show what the error looks like looks like in Audition? I get about -69dB RMS difference with a single error peak at around -35dB around the 25 second mark.
I see a greater error if I just subtract these without removing a small clock drift. Which brings up an interesting question: why did the clock drift between these at all if what you changed were two analog cables?