Recently published papers have proven the effectiveness of Electromagnetic time reversal (EMTR) in locating the position of faults such as phase-to-ground shunt connections in power grids. EMTR directly transposes the idea of focusing energy back to its source introduced in original time-reversal methods. The current interpretation of EMTR, based on metrics measuring energy or peak-amplitude of focusing, is shown to suffer from ambiguities that increase the risk of inaccurate fault location. After pointing out under what conditions time-reversal focusing occurs, an original frequency-domain reformulation of EMTR is introduced, showing that EMTR should more accurately be interpreted as a correlation estimator. New metrics based on this observation are introduced, taking into account the inhomogeneous transmission of electrical energy throughout complex networks, enabling a direct quantitative evaluation of the likelihood of locating a fault. Extensive numerical simulations confirm that the proposed formulation systematically improves the reliability of EMTR location estimates when faults occur in power grids of realistic complexity, highlighting the accrued risk that comes with the use of metrics that measure the scale of time-reversal focusing rather than its quality.