If I spend more than 50% of the time in an interview solving puzzles I'm not working for you. I may suck as a programmer, but if that's what you do in an interview, you suck at interviewing people.
Puzzles in interviews are worthless unless the potential job involves doing those same kind of puzzles. If you want to see how someone's mind works, you could, say, talk through solving a real-world problem as if you were pair-programming. But don't use puzzles are brain-teasers. They don't tell you anything practical.
If the puzzles are "why are manhole covers round?" I agree.
If the puzzles are "given a list of N integers, positive and negative, find the consecutive subset with the most negative sum" then I want to see how they come up with an answer.
