The hypothesis that people on average will cut the calories in their orders when the amount of calories is labeled looks very non-obvious. I can see several opposing factors that might be in play when the restaurant starts to make nutrition information available:
I think it's presumptuous to conclude that labeling the amount of calories in food "totally fails to change people's eating behaviors at all", but rather that measuring the average calories in the orders is not enough and more thorough investigation is needed.
"The hypothesis that people on average will cut the calories in their orders when the amount of calories is labeled looks very non-obvious"
It may be, but nonetheless, reducing obesity is the justification for recent restraunt labeling laws.
Information that surprises you is interesting as it exposes where you have been miscalibrated, and allows you to correct for that.
I suspect the users of LessWrong have fairly similar beliefs, so it is probable that information that has surprised you would surprise others here, so it would be useful for them if you shared them.
Example: In a discussion with a friend recently I realised I had massively miscalibrated on the percentage of the UK population who shared my beliefs on certain subjects, in general the population was far more conservative than I had expected.
In retrospect I was assuming my own personal experience was more representative than it was, even when attempting to correct for that.