The craters are certainly still a work in progress. That being said, about the flat interiors of the craters I will say a few things.
1) It seems to be in line with photos of lunar craters, esp. the large ones:
http://www.madpc.co.uk/~peterl/Moon/Cra ... nines.html I expect this is because the larger impacts essentially liquefy at the point of impact, and so level out (but I am no lunar geologist, so that may not be correct.)
2) What you think "looks good" may be largely down to what you're expecting to see. I've spent quite a lot of time lately looking at lunar craters, so that flat interiors of the craters don't seem too out of place to me, at least in the case of the larger craters.
3) The craters actually do slam a bowl shaped depression into the height map. However, the height map is of limited precision (1 bytte in the current implementation) and if the terrain is already close to zero (likely) then it can and often does "bottom out" ate zero, producing the flatness that you noticed. Fixing that is probably a matter of tuning the algorithm a bit to get a height map that's not naturally so close to zero.
The main complaints I have so far with the craters are:
1) the height map perturbations are tloo large scale compared to the terrain generation height map perturbations, resulting in clipping and or overflow at 0 and at 255.
2) Not nearly enough small craters.
3) The colors are determined by altitude and by position in a 3d noise field to offset into a color map. Sometimes this works well, giving the "rays" a contrasting color, but sometimes it doesn't work well. It might be neat if the craters actually did some kind of "color displacement". Then again what I've got now might be good enough, esp. if I can get the normal maps working.
4) Probably should only have "rays" on the larger craters, not on every crater.
So yeah, I am not totally happy with the craters, but neither am I totally unhappy with them.