Compression mystery - Thought I understood but still baffling

Dear all,
My issue is based on my understanding that the higher the ratio, the greater the compression and therefore the greater the reduction in gain. So if this basic assumption is wrong the rest of my post will be silly to some.
 
I have a drum loop and I am experimenting with compression values. I am using the built in PC4K bus compressor in Prochannel - but using it on the stereo track.
My fader is at "0". The drum loop is playing with no effects or alterations added. The loop peaks at -1.6db
I am playing with compressing only the highest elements of the sound. My threshold is set at around  -17db (-16.8 to be exact). The ratio is set to 2:1. Attack is set to 1ms and release set to 0.4s. It is fully "wet" and there is no makeup gain.
 
At 2:1 ratio the gain reduction seems to be around 6db. The peak meter on my fader reads -3.3. So although the gain reduction is 6db this brings down the level of the highest peak by about 2db.
 
With the ratio at 4:1 but everything else left untouched the gain reduction drops to around 4db and the peak meter now reads -2.3. So with a heavier compression I only get 4db of gain reduction and my highest peak is still creeping up.
 
With the ratio at 10:1 the gain reduction is just under 4db and the peak meter reads -2.3 again.
 
I thought higher ratios meant the signal was squeezed more tightly. Why do I get higher gain reduction with lower ratios?

Scook...Thank You!! Much appreciated.
Having read those threads it seems I am on the right track in terms of my understanding of what the compressor should do. Unlike some of the other posters in those threads, I am not hung up on whether the PC4K actually does it in the pproper way, or if it is out etc...You don't calculate first then apply compression. If the PC4K has its quirks I can live with that. I was just afrais my whole understanding of compression was wrong.
 
Also, Unlike quite a few posters, I really do like the ProChannel, its EQ and the effects. They are clean and useful.
Thanks again Scook...saved me a lot of hours searching and re-experimenting.