Obviously there are situations in which a pitcher's ERA doesn't match his DIPS and there' something to it. Buehrle's ERA was lower than his FIP, people looked into it, and found that he's an absolute elite defender and that makes up most of the difference. Matt Cain had something going on. Knuckleballers and guys with elite changeups tend to give up fewer runs than DIPS suggests they should.
I think what frustrates a lot of people is the IP threshold at which you start to take a serious look at it. Half the forum is calling for Aaron Loup's demotion even though he's been outstanding because they're looking at ERA in a 30-inning sample. People couldn't stop praising Sanchez because his ERA was shiny over a 50 inning sample.
There's a time to start taking a serious look at ERA-xFIP discrepancies. That time isn't after 50 innings, and the evidence suggests it isn't even after a few hundred innings.