Great video! Thanks for that -- it shows it clearly.
The composite mod monitor
looks smashed because of the extreme amount of clipping. If you look at the signal on a scope, it looks like it's been cut with a razor blade. However, it doesn't
sound smashed or distorted because thanks to my psychoacoustically distortion masking clipper, clipping distortion is masked, so I don't need to do any HF limiting before hitting the clipper. It's essentially the best of both worlds -- the punch and impact of a clipper, but the spectral purity of a gentle limiter. People listen to L/R, not MPX, so the L/R mod monitor displays are much more indicative of what it sounds like -- clear high-end with actual dynamics.
If you de-emphasize the audio before looking at L/R, the meters do not move nearly as much, however de-emphasis doesn't remove dynamics, it just changes the frequency balance, so all the HF dynamics remain after de-emphasis, even though they'll be obscured on a wideband meter due to the then much louder lower frequencies.
Bojcha is absolutely correct that it's only the highs that go over 100%, and that comes in
really handy when you're fighting a pre-emphasis curve.
Doing this same test on most processors you'll see the MPX meter move more than the L/R meters (the ones carrying the audio we're supposed to be listening to), which is the exact opposite of what we want

.
Best regards,
Leif Claesson