Well, first of all: Taking the "
thump away" i.e. controlling/softening the transients is one of the main task compressors are built for.
You can make the softening effect less "obvious" by increasing the threshold, attack or reducing the ratio. You can only reduce
the damage to a point where the compressor does nothing, which brings us to a valuable idea: Don't use compression without a good reason.
A good reason may be the need to "even" out a performance or pushing the low level areas by a few dBs, etc. A compressor introduces particularly bad side-effects, it has a serious hit on the crispiness, "punch", musical integrity, noise floor and countless other things.
You can't increase the punch with a compressor, that's a misunderstanding! What you can do is totally squeeze the sh!t out of some music with unreasonable ratios, but keep the attack open is such a way that the small initial transients pass through the compressor with slightly less damage. But that doesn't create punch, and proper equal loudness bypassing will clearly show that this technique doesn't really work: The original material will sound crispier.
If you need more punch, select sound that has punch! or record them with punch! Or program a synth in such a way that it create punchy sounds. Then, you can start thinking about intentionally sharing a little bit of the punch for a little bit more cohesion and density (via compression). In this case, the result will be loud and fat, but still very crispy/punchy. That how great sounding records are made. Processing can only control/restrict the original material, but never truly enhance it! You can't polish a turd for beauty.
On a last note, have a look a transients designers, these are much a better choice for the task.