As far as I understand, one should get the strongest signal ...

As far as I understand, one should get the strongest signal possible, keeping your sound as close to zero as possible without clipping. But then what happens when you have a peice that is pretty loud to begin with, and that peaks at one point. According to this Rule of thumb the peak would be close to zero and the rest of the piece would be under leveled. Doesn't it?

In the olden days of tape, you could actually go into the red a little and not burn the recording. With digital recording, once you clip it, it's toast.

So yes, you'll want to record the signal as hot as possible without clipping. Softer passages will be quieter, it's what we used to call dynamics.

To help cure a wide range of unwanted dynamics we introduce the compressor. It will take those real loud sounds and "Compress" them down a bit so they fit into the dynamic range. Audio engineers used to fly the mixing console by hand and lower the volume manually when the orchestra started to swell.

Older compressors used to make themselves known by bringing up the gain during soft sections, you could hear them swell and squish on the recording. Bleh. They've gotten much better.

So, with all that history. Here's my rule of thumbs. I record with no or minimal compression. let the dynamics happen. I'll use the compressor to catch something that would have ruined the recording but I don't let it run the board for me.

Then in post, once I have a good clean signal, I might at a little more light compression to keep the signal within a dynamic range. I don't want people in the car fiddling with the volume knob if they don't have too.