Alright, so I've done a bit of digging. If you're interested, you may want to review this good primer on the
Microsoft WAVE format. It appears as though the MT2496 - like the JB3 - adds some useless sub-chunk data to the WAV file. The MT2496, in writing the WAV header, does not appear to take into account this extra sub-chunk it adds to the file. So we're left with a WAV header file size descriptor that does not match the actual file size (i.e. the WAV header file size descriptor is always
smaller than the actual file size). Hence, CD-Wave (and perhaps other apps) warn that the file size as reported by the WAV header differs from the actual file size.
The message FLAC presents (please note the contents of the sub-chunk data may differ from one file to the next):
WARNING: skipping unknown sub-chunk '⌠5☺;'
The specific message CD-Wave presents:
File header reports a size that is different than the system reports. Do you want to adjust the size? (Yes to override header, No to trust header, Cancel to abort)
What I've found is it doesn't matter how you answer, Yes (override header) or No (trust header) - CD-Wave strips the sub-chunk data regardless. Likewise, CEP strips the sub-chunk data. So does FLAC. Haven't tested Wavelab, but I suspect it strips the data, too. And one may also strip it manually with hex editor. Basically, I think this is not something about which anyone should worry.
But..."Ack!!", you might say, "what happens if my app trusts the WAV header file size descriptor and therefore doesn't handle any data beyond the file size as designated in the WAV header?" I think that since the MT adds the sub-chunk data at the END of the file, the only thing that will happen is the app will not handle this useless sub-chunk data. In other words, nothing happens, other than the useless sub-chunk data is stripped from the file.