You all make some very good points.
Allow me to show you how I see it. I work with numbers, I was very good a physics and mathematics in my studies. I know that was nearly a decade ago, but I remember a few things. As long as you have the raw processing power dedicated to it you can simulate just about everything, even the human brain. When the raw processing power is limited and not infinite, some things have to basically be chosen as arbitrary and not processed. If you want, educated guess work. It's more that obvious that since FM to FM2 and FM3, by squeezing more processing power out of the XBOX360, there have been more and more routines added and less and less of those educated guesses. The sheer dedication to improve in the physics department so much is something I commend T10 for, especially given the relatively short development cycles compared to PD. However, there will always be flaws, it's a simulation, so sooner or later someone with the infinite processing power of the human brain, that trumps any computer, a player, will find ways to bend those rules. We're not talking Neo here, just someone finding a loop hole in the physics that others missed. Some of the loop holes were actually patched not long after the release of FM2, if anyone here remembers. That's how some have managed to get unrealistic lap times. Yes, the simulation of physics did have minor flaws, and to achieve perfection on a console would be a very unrealistic demand, even on a supercomputer it would be difficult. However, T10 are making a mathematically correct effort to get as close as possible as the hardware allows, to reality, which is more than I can say for the physics of PD. Why? Because if GT5P or indeed GT5 had something as good or better, they would have also boasted about it. Instead they boast about a near perfect car models and and motion captured pit crews or spectators. Got to give them credit for the track design though, far closer to the real life circuits than FM2.
OK, that may make some heads spin around here, sorry, I went a bit deep.
Now to the "all cars over steer in FM2" statement. I believe it to be false. Under sustained heavy power load a lot of cars in FM2 tend to under steer as the car simply can not swing the mass around the corner no matter how good the the setup is. Yes, this does happen in FM2. Also AWD cars in FM2 do under steer quite badly at times, as they do in real life. There certainly is over steer in FM2 and, this is just my experience though, a lot of cars over steer, it's very common. In my opinion the over steer and under steer balance is far more realistic in FM2 than in GT5P. Though, as some of you said, GT5 may have improved further in the physics simulation department and I'm sure FM3 has too...
Lastly, yes, comparing lap times to real life it's not really going to mean the simulation with the closer time to real life is the better one or the more mathematically correct one. A poster here put two videos up, the times were 4s apart and, as someone's already mentioned, the FM2 track was wider and that allows for straighter lines to be taken through turns. Let's not even take into account that there is more than one race tyre in FM2 and that the drivers, or indeed driver, may not have been identically poised on both laps, is using different controllers most likely, looking at different displays with different refresh rates and, of course, playing on different consoles. I could go on but hopefully you lads are getting the point right about now.
Really? You pulled that one out? 
