It's lighting and overall look is just more realistic than any racer on the market.
I will not dispute your claim as far as "overall look" goes, because that's your opinion. But as far as lighting goes, PGR2 is better than GT3, both technically and artistically. But then again, PGR2 is
newer and thus
all of it's lighting techniques have the benefit of being more researched and realistic.
This has to be realized, because I expect PD to
improve how their lighting is done in GT4.
Also, PGR2 has the benefit of higher polygon counts, which make the vertex lighting less cheap looking, as it does on in GT3. In GT3, you can see all the triangles lit up on the roof/hood of the car, while in PGR2 it looks like one solid mass and it's lighting doesn't look so cheap. To lightmaps in PGR2 apply to nearly everything in the scene. When you look at the buildings, their ledges, window-sills, anything that extrudes casts accurate shadows. My only nitpick is that the railings (added to the map for each race path) and lightposts do not cast shadows. Trees also do not benefit from this raytraced lightmap, so the artist add shadow textures to the road. Have you noticed that these lightmaps are crazy accurate? The raytracing is crazy since it takes into account that if you are between 2 buildings and none are lit by sunlight, that the shadows will be darker. This also applies to under bridges (but that's obvious). It also applies to buildings too, and the fact that it's not simply a one-tone shadow raytraced out of shape, but also out of lighting is simply mind-bogglingly awesome.
Now this all comes back to my earlier comment about the right tradeoffs. Obviously, you don't notice the lighting that much, while I (since I've modelled/programmed 3D demos) can see it clearly. If most gamers are in your position, then PD made the right tradeoffs.
I regularily play 6 player GT3 with my Comp Sci friends, and they notice how much superior PGR2 is to any other game. There are many times where our jaws collectively drop as we view the replays. Judging from this, I think Bizarre made the right tradeoff. At 30 fps, this game is a graphical achievement, a benchmark. At 60 fps, it would have looked like any other game. On the other hand, if this game didn't control at 60fps, it wouldn't have been much of a racing game.
Personally, I think the artists at PD have the right idea. They don't have the benefit of all the lighting and tricks the Xbox can pull off, so they rely on smart texturing. I feel that they have the ideology that "unless we can actually do pixel shadows and hold large raytraced lightmaps in memory, let's just fake it". I think it's the correct way of thinking.
We should be comparing games like SEGA GT, and PGR2, because in a lot of ways, SEGA GT represents what PGR2 could have been at 60 fps. Most GT3 fans I know are not impressed with SEGA GT. Yes, it's at 60 fps, but it only has a few good things going for it graphically, and the artists didn't do a great job of capturing a realistic look (while techically, the models are great, effects like the blurring and specular highlighting are fantastic), overall it doesn't do much more than GT3, and as always, the texturework in GT3 is better. GT3 fakes detailed lighting with it's texturework, while SEGA GT purely relied on realtime lighting. Throw in pixel lighting casted from objects that can be applied to itself, and SEGA GT will look a whole lot more realistic...but the framerate is gonna take a hit, no doubt about it. So when Bizarre went the route they did, I felt that they made the right choice, otherwise their game could have looked so much like SEGA GT, but with great backgrounds.
