[abiglan]

Depends on many things, but if the data appears "controversial yet conclusive" then you investigate more deeply until you have ruled out just about anything else. Long time ago, I was part of a research team that sat on a 3.5 sigma result that only got stronger when we got more restrictive with the data set. We ran for two more years looking for other explanations. I was in grad school and could only think "why don't we publish?!?!" but the more experienced folks won out and upon further review, no Nobel Prizes were awarded and the reputations and funding remained in place. I think the adage "you only get to cry wolf! once" applies here.