Excellent link. A particularly noteworthy excerpt:

[Q:] I assume that most people in these jobs aren't actually trying to convict innocent people. So how does such misconduct come about?

[A:] I think what happens is that prosecutors and police think they've got the right guy, and consequently they think it's OK to cut corners or control the game a little bit to make sure he's convicted.

This is the same phenomenon that is responsible for most scientific scandals: people cheat when they think they have the right answer.

It illustrates why proper methods really ought to be sacrosanct even when you're sure.