yeah, the compression only kicks in when the threshold is exceeded, so in that sense it is kind of a gate.
The make up gain doesnt bring the max peaks down, thats the job of the compression (the amount of compression being controlled by the ratio). The make-up gain is simply a volume knob, I'll try and illustrate what goes on:
Imagine you have a track that is 15db for 5 secs and 25db for 5 secs. Your threshold is at 20db and you have an reasonably large ratio giving some pretty heavy compression. This means that for the first 5 secs, the track is the same, but for the next 5 secs it is compressed. This reduces the overall apparent volume of the track it's still 15db for 5 secs, but because there is quite a lot of compression on the next 5 secs, it is something closer to 20db. The make-up gain is there to push the volume back up, but it affects the whole track. So pushing it back up so that the last 5 secs is back at 25db will also push up the first 5 secs by the same amount (to 20db).
Now, before you say that 'hey, why dont you just automate the first 5 secs so it is louder', remember that the compression has shaved off the peaks of the loud parts and in doing so has changed the sonic character of the track.
In that respect, compression can be seen as having 2 functions - it can reduce the dynamic range of the track (like you said - make the quieters seem louder by reducing the loud parts and increasing the volume on the whole thing), but it can also be used subtly to change the sonic character of a track. It takes some playing round to get used to that, and really I am still a novice when it comes to that. For the most obvious effect, putting compression on a kick drum can make it 'pound' more or give it more 'punch' or give a snare more 'snap'. It can tighten up guitar tracks and make bass warmer or more pounding again.
It requires a lot of experimenting with primiarily the ratio and attack and release times, and to a lesser extent the threshold and make-up