Then you must modify the experiment in order to support the theory. It won't be valid if it the rest of the information contradicts the theory. Either the theory is incorrect and must be modified to fit the experiment or the experiment doesn't support the theory at all.
