You can also have initial dynamic peaks that register on a peak meter, but the initial peak is too quick for some types of compressors to start reacting. It starts indicating compression when you increase the input because more of the "average" signal below the initial or higher peak causes the compression to kick in. So you end up shoving the peaks too high in order to make the compressor react to the part of the signal that has a enough higher average or rms value to make the compressor kick in.
This is not uncommon in "character" compressors at all. Many vari-mu type compressors act this way. The EMI RS124 even has a hold function that requires you, in some cases (depending on the source), to hit the compressor with the proper signal level, and then hold the compressor at that amount of compression while the track is restarted, until the initial peak kicks in. Then release the hold so the compressor starts reacting properly to the signal. If any long drops in signal occur within the track, the compressor must be put into hold again until the signal comes back in.
You can try this yourself with a compressor using the attack, release, and threshold settings. Just increase the attack time. You'll find it takes an increasingly lower threshold (or higher input) to get the same amount of compression, the more you increase the attack time. The initial peak is too fast for the compressors attack time. This will all vary depending on the source and its dynamics.
Hmmm... now that I look at the PC76, it does have an attack setting. So you still have this problem using the fastest attack, and an appropriate release time for the material?
I don't know if this is the case with the PC76 (I never use ProChannel). If it is, use a peak limiter just before the compressor so that it clips off the quick overs, but the threshold of the peak limiter is too high to affect any of the rest of the signal. Then only the PC76 would compress the rest or majority of the signal, and add its "character" to the track.