I got a R&R at a pretty good journal. It is an empirical paper. The most substantive comment was that I should flesh out the model more to make the contribution more clear. I expand the model fairly substantially, and run the empirics, and they are consistent with my model. Great.
Then I find and fix a bug in my code. The results are no longer consistent with my model. They are however consistent with another model, which has been published in AER.
My question: does it make any sense to retain my model, and then conclude "nope, actually my model doesn't work. Use the model of Authors et al. instead.". I don't suppose that this would be enough for a second-tier journal, would it? Unfortunately, I don't think that the refutation of my model would be as interesting as evidence in support of it.
Advice?