If you are modulating at 20 or 30 KHz this could be outside the bandpass of the matchbox. This will present a different load on the rig and maybe causing a high voltage to be generated in the network because it isn't going to the antenna. I bet you notice the swr change with modulation.
So the way to test for that would be to limit the audio passband?
Strap in a carbon mic or such?
That's an interesting point as I used to have this problem with my T-3, guess
I could have tested the theory by squashing down the audio with the EQ?
Now that I've got my old audio processing gear up from Buffalo, I'm tempted to fire up the T-3 again.
-Bruce