My thinking was that if one assumes that there is a function: calories -> life expectancy, all else being equal, then the normal intake may or may not correspond to a local maximum. If it does not, then mild starving ought to increase life expectancy a bit. Of course, it tells us nothing about the global behavior, including significant restriction.

Oh, and I'm not sure what you meant about upvoting/downvoting. Not that it matters.