As much as you save in one way or another, the recognition problem is still there.
Perhaps (my understanding is that you started this issue report by saying the issue was not seriously occurring in recordings), but if you give me a SaveThatWave file, I can review the actual OpenEars session and what OpenEars recorded using its audio driver. On the other hand, if you send me a file that was recorded with a different app, I won’t know anything about what OpenEars’ audio driver did in the session. And if you don’t accurately tell me how an audio file originates, I will spend a lot of time looking into something without knowing what it is, which wouldn’t bring much. It’s going to take a while to get to review this while the next update is released, so if this is a showstopper for you and you have time pressure, check into the alternatives as well.