It shouldn't be dismissed. Suppose two candidate hypotheses are (a) no dark matter and (b) enough dark matter to produce an expected signal at +1sigma in this experiment. Then a 1.5sigma event is about [EDIT: following ratio was wrong, sorry] 3:1 Bayesian evidence for (b) over (a). Of course that's nowhere near enough to justify saying (b) is right and (a) is wrong, and of course the simple normal-distribution model I've assumed is likely too simple; but I think a 1.5sigma event should certainly lead you to adjust your probability estimates a little.