So is this a completely different outcome in the second analyses or is it partially related to the first. When moving forward you still need to think about false discovery rate and trying to maintain your original level of significance (alpha), because after awhile you may find significance by chance (e.g., 1/20, type I error for an alpha=0.05).