To me, the whole point of Headshot is: I give you two images, and you create the character. At the moment, this doesn't seem to work very well. I've attached a perfect example of what I mean. The head shape and nose are completely wrong.

If this was free software then I wouldn't mind but, given the price, it should work as expected...especially on its third iteration.

If the problem is in the source images then, given that you have added AI to the software, you should be pre-processing them as required. In fact, since you are using AI - why not let the AI generate all the character settings from the input images?