When you cannot refute, ignore it. Wasn't it the standard practice so far? (Mentioned in the paper.)

Instead keep the fancy unexpected results published and cited, even if irrelevant or wrong.