I think the points that you bring up are mostly true for well established and common platforms and architectures, but the author, Mike Acton, focuses mostly on the PS3, where it is more conceivable that a compiler would not e.g. optimize a division operation to a shift, especially for an SPU.
I don't think it's fair to generalize game programmers for premature optimization nonsense. Game development certainly has much stricter requirements on performance than say, Joe's Ruby on Rails website, and so I would be much more inclined to trust them for optimization tips. That being said though, yes, a lot of them have no idea what they are talking about.
I have been on both sides of the fence. I have worked with people who attempt to optimize every single instruction, and with people who gladly misinterpret "premature optimization is the root of all evil." Both of them are obviously toxic in their own ways.