It is certainly amazing to see just how closely the predictions match the results. That means the theory is really good. However, the finding that the actual value is off, and reliably different from the prediction, even by a small amount, certainly suggest the theory needs tweeking.

It's like when Carpenter told Professor Barnhard that the prediction of the first term was essentially the answer. Barnhard asked about the other remaining terms, and Carpenter said their contribution wasn't significant. Barnhard then asks Carpenter how he knows...