It basically balances a previously-mentioned theoretical risk by explaining why the outcome predicted by that theoretical risk differs from the actual, observable outcome.