Right, but what I'm saying is that the third person shots aren't rendered in game. The background is, but Amanda is rendered separately and then composited into the shot.
I guess we'll never know for sure what went on, but continuing to speculate for the curiosity value...
I had another look, and I'm not sure they are composited. Are you thinking this is the case from this quote from the makers?:
The final series is comprised of three types of scenes: brand-new scenes that are rendered from scratch, cinematics taken directly from the game, and first-person scenes from the game we re-shot, edited and inserted a CG model of Amanda Ripley.
I think this is a rather ambiguous statement, and could just as easily mean they converted the 1st-person scenes into 3rd-person scenes by going back into the game engine but his time including the (low-poly) Amanda model.
ON THE OTHER HAND some of the shots DO look very odd (as though composited on top):
In this top down angle of Amanda walking past Ricardo, look at the dodgy shadow she's casting on the floor. Not only is it very low res and lacking in density, it also sits on top of the existing shadows being cast in the room. Surely a sign that she is comped on top, yes?
But then again, in this shot, the shadow from the clamp laying on the floor seems to fall on her foot, which would be very difficult to achieve if she was comped:
Plus, sometimes she crosses behind other objects in the scene. Not a problem if you were willing to rotoscope an animated matte around (in this case) the chair backs, but I don't think they had the budget for that kind of messing around:
This actually happens quite often:
But then, going back to the scene where she converses with Taylor early on in the movie (and probably the most egregious shot of all), look how horrible Amanda's close-up is integrated into the background, exactly like she was poorly comped (you can even see a dark edge around her highlighted hair):
There are also many shots where they have added a random jiggle to the camera to simulate a hand-held effect. When Amanda is in the foreground, it creates a disturbing "swimming" effect, which is normally a giveaway of an imprecise 3D match move. IOW, they had a background (without Amanda) which had lots of camera movement, and tried to add a new Amanda to the foreground, having to duplicate the camera movement in her element too, but not quite getting it right.
BUT, this same swimming effect can also occur if you don't use the right sort of jiggle on your camera; namely, you want to make your camera jiggle through rotations, not translations.
What does this all mean? Danged if I know...