Skimmed that but didn't read that in full -- I'll get around to it. Thanks a lot.

Isn't it roughly what you'd expect, though? The authors are saying that the IAT isn't great at predicting discriminatory behavior. But discriminatory behavior depends on more than just your biases; you will discriminate more if your environment doesn't punish you, for instance. That kind of environment/self-interest factor might matter more than subconscious bias. (Your likelihood of, say, hiring a diverse workforce will depend more on the local population, the local laws, and the industry you work in, than your personal biases.)