That batch file certainly soundslike an interesting idea. That might be one for Fluesopp to implement though.
I have an idea that might make jittered trees stay in the alpha mask area more. It shouldn't be too hard to implement either. It might slow the program down even more though but meh. Waiting a few more seconds is still a LOT faster than spending hours on a single forest.
Btw, it will be the accumalative jitter that is making the most change to the trees going over your mask area I would guess. The other jitters just move the trees a small amount relative to the size of the tree. The accumalative one moves them relative to the size of the density and has a knock-on effect with the trees in the same row/column. It's meant for forests where the density is turned down so you get lots of randomness there.

You mentioned in an earlier post that you thought the generated forests were not coming out as square as they should be. Or I guess the same as the aplhamask? Is this still the case? Or is it this jitter that is making them appear out do you think? I ask because the statement has been bothering me and although the shapes seem ok to me, there is obviously something wrong when you use taller trees. The shape seems ok still but you need to leave quite a large blank area at the top of your mask or trees seem to get cut off. I'm guessing there's a problem with the scaling somewhere but I'll need to think on it.