Thanks, I've had a go at this sort of thing, so I'm quite excited at the prospect of a mainstream game adopting such a risky strategy!
GT5P was the first GT game to get a "sound simulation" credit, and it was mostly to do with how the sounds we're mixed, processed, filtered etc. according to the physical principles that govern the real thing. There was a bit of physically correct synthesis going on, too. AES is more like a continuation of that approach, another piece of the jigsaw.
Acoustic transfer is all about how structural vibrations "transfer" to the air (where vibrations are called "acoustic"). I use it more generally to mean mapping the 3D propagation of sounds from complex sources, be they structural, acoustic, whatever.
Most games use monopole sources, the simplest imaginable kind (perfect spherical radiation pattern), and theoretically impossible: unphysical. They work well with mono recordings, because they add no extra colouration, which would otherwise clash with the colouration already in the recording. That colouration comes from the mic placement (and its type etc.) relative to sound sources and the environment, i.e. where it sits in the resulting (real) 3D sound field and its hugely complex interference patterns.
Codemasters use a mixing method that captures the virtual 3D sound field at the listener's location, but only if that sound field had been created by monopole sources. It's better than the unphysical "panning" in the usual multichannel mixes, but the transfer is still more important because it yields more realistic interference patterns at the listener, which would be obvious even in a mono mix down.
Generated sounds don't have any colouration, so the use of more complex source models does not reduce the quality; in fact it is absolutely necessary to guarantee quality!