Knowing nothing about the code, I would agree that a to-hit chance would be awesome. (My own 2D space sim design incorporates that to try to reflect the sigmoid curve of firing at the centre of mass and having shots go high or low in the theoretical 3D game environment.)
However, knowing how the code is all about building bounds polygons, this would grow extremely complicated extremely quickly. Instead of collision-detecting against a single mass of triangles, what it would then be doing is breaking down the structure into a number of "bounds layers" and then running collision detection against each. While in essence the number of bounds polygons is unlikely to change, the overhead incurred in having to check against multiple layers would result in slowdown -- paired with the pseudo-random number generator and it incurs even more overhead.
It's one of those things that'd be tough to make a ruling on without having actually implemented it, and after having implemented it the work has already been done and the point is moot. I'd lean towards "sure, why not?" just for sake of liking the idea of building a ship with struts that don't actually reduce the hit points of the hull itself -- Uoomoz's Corvus' McHammer would benefit immensely if the strut and hammer behaved only for collisions and the primary hull itself was the only thing that projectiles could affect, and it'd make the AI smarter about not wasting shots against the hammer when the hammer is just swung out of the way. That example does raise the question, however -- are these "semi-permeable polygons" still guaranteed collision points for ships? What about missiles? And how does the game handle "duration" while remaining in the polygon's bounds from simulation frame to simulation frame? Will it spontaneously react to a collision eventually or will it be first-time's-the-charm?
My point is basically that while this seems simple and elegant on the surface, it raises a bunch of border scenarios that rapidly grow it in complexity, like mercury thiocyanate. ;-)