I'm assuming that you refer to EPC values that are reported as part of the modification index output in Mplus. The EPC tells you what the parameter estimate would look like (approximately) if you were to add the parameter in question to your model.

When variables load on multiple factors, and the factors are correlated, standardized loadings can be > 1. A value of 1.07 would therefore (potentially) represent an admissible (i.e., not necessarily improper) loading estimate. It is interpreted in the same way as a partial standardized regression coefficient in a multiple regression analysis with correlated predictors.

Christian Geiser Thank you very much for taking the time to provide such clear and helpful guidance.

Kind regards,

Emilie