interesting! my sense with the MF measures is they tend to have nicer distributions (in part because they are bounded between 0 and 1). I'd be curious if MF measures do better in your dataset or not. The non-correlation with self-report is so interesting though!