Sometimes, the results of an experiment contradict what we have strong theoretical reason to believe. But experiments can go wrong, for various reasons. So if our theory is strong enough, we should in some cases defy the data: know that there has to be something wrong with the result, even without offering ideas on what it might be.