Yeah, but you don't need to postulate different definitions of sex then; they're fudging the numbers, and nothing tells you whether they use this parameter to do it. I was proposing a model where people are completely honest and still give different answers.
I read in Cosmo (probably the least reliable source possible) that women often don't count one-night-stands. No idea how to test.
Different definitions of sex sound like a plausible way of fudging the numbers (I don't see any other ways of fudging the numbers that aren't outright lying), and fudging the numbers seems like a good motivation to have different definitions of sex.
The lie detector in the OP makes outright lying seem a bit more plausible though, if it's true.
From The New York Times:
So how sketchy is the research on human sexual behavior, anyway?