Finding the Point of Impact for an Impulse Resolution

I've been building a small racing game in C#, and I've hit a bit of a block when it comes to the impulse resolution of the car to car collisions. At the moment, the cars are able to hit and push each-other, but there's no resulting rotation from these collisions.

The problem I'm having is determining the point of impact.

I'm using a set of Vector2s to represent the outline of the car. From this, I'm using SAT to determine if there is an collision. While looking for a separating axis, I'm also keeping track of the axis of least penetration, this is the axis that gets used as the normal for the impulse resolution when there is a collision. The axes that I'm using are the normals of the sides of the cars that are being checked for collision.

The collision detection part of this is working very well; the way that the cars are able to shove other cars out of their way isn't too bad either. But, I want the cars to be able to spin out of control when hit at an odd angle.

The best solution I've thought of so far goes something like this:
* CarA is set to the car who's side is being used to determine the normal for the impulse resolution.
* CarB is the other car.
* I push the cars apart to get rid of any penetration (they will still be touching, just not overlapping)
* I use the point of CarB that is nearest to the impacted side of CarA, as the point of impact.

As far as I can tell, this solution seems sound, but I'm being bothered by one thing. What if it's a flat side-to-side hit? A perfect head-on-collision of two cars playing chicken would be a good example of what's bothering me.

Does anyone have any advice on what might be the best course of action for this?

Tags: C#, Game Development, XNA, Simulate, Collisions