This must be cherry-picking stats on both scales -- what percentage of offenders against children are pedophiles, and what percent of men are pedophiles. A critical variable is the cut off age of "child" -- if it is 17 then real pedophiles would likely not even be tempted.

Hall, Hirschman and Oliver found something like 25% pedophiles but the scientists I've talked to think it's an outlier. I've been using 1-2% for strong preference to prepubescents over adults on the one hand, and more like 10% for "aware of significant attraction to children".