I work in Germany now. I believe the code was originally developed at Aachen, and BMW, Mercedes, at least HAD copys of them BMW have since gone over to GT Power as have Porsche and Merc. I don't know how much BMW uses it now. You're right in that it is a Finite Volume code (but I don't think it's FRAM or Lax Wendroff), and not Method of characteristics and I know it has also been used succesfully in NVH prediction work, which makes me wonder whether it is centre biassed in ressolving the spurious gradients. I don't know if it uses a Collocated Grid.
I've heard it uses a look up table for junction losses, but that's not what you asked!