I had a look at the descriptor info for the CM6533.
It has a setup where it mixes microphone input (from the MIC jack of the device) and USB input to produce sound on the output.
My suspicion is that USBAUDIO.SYS does not do the necessary setup to the FEATURE and MIXER components (look at the descriptor info) to ONLY take USB input (and turn off MIC input) and route that to the sound output. The problem is that the "sound network" between inputs (be it MIC or USB) and outputs (be it SPEAKER or USB) is "universal" and the connections a bit difficult to parse and come up with a generic decision on how to set the FEATURE and MIXER components (maybe someone DOES want to mix MIC input and USB input to a common speaker output).
"Cheapo" devices usually have a very simple network: connect MIC input via FEATURE component to USB output, connect USB input via FEATURE component to speaker and that's what USBAUDIO.SYS (and USBAUD2.SYS) is aiming for.
(short note: a FEATURE component does things like mute on/off, setting volume, setting base, setting treble etc. Your device only allows for mute on/off and volume as most devices do).
If I find some time, I would be able to hack a REXX script to properly set the FEATURE and MIXER components to do the above. It might take some trial and error though, as the REXX interface (using USBCALLS.DLL and USBRESMG.SYS under the hoods) is a little difficult to use.
You would then at least be able to run the REXX script shortly after you plug in the USB device. The USB device itself will keep the settings until you detach it.
One way to check my suspicion would be to attach a MIC to the MIC jack and then see if the picked up sounds are being heard on the speaker while you are playing that WAV file. One the other hand, that would not explain why you can successfully play other WAV files without any probs ...