Mutation-based fault localization (MBFL) is a new promising direction toward improving automated fault localization by leveraging mutation testing results. One issue in experimenting MBFL techniques is that MBFL results are more prone to validity threats because there are many factors to control in conducting mutation testing for fault localization. To understand different validity threats in experimenting MBFL, this paper reports our reproduction/revisit of the MBFL assessments with Defects4J with different controlled set ups. Especially, to eliminate the existing noises, we systematically reviewed and revised 26 Defects4J artifacts, From the comparison with the new and the existing experiment results, we found the configuration of the controlled variables change the results of the experiments in a large degree, therefore, these threats to validity must be carefully limited in emprical assessments of mutation-based fault localization techniques.