Actual use of Statistical Energy Analysis for complex systems requires the numerical or experimental determination of the SEA loss factors (E-SEA). This approach involves experimental and numerical issues that can lead to results inconsistent with the SEA hypothesis, namely, the SEA loss factors obtained may be negative. This work presents an analysis of the E-SEA energy matrix class that shows that it is not well suited for this type of inverse problem. In addition, a method based on Monte Carlo Filtering capable of overcoming this inconsistency is presented, so that robust values of SEA loss factors can be obtained from experimental results. The method is described and applied to two application cases to show its practical implementation, the advantages that the method can offer and its comparison with other methods presented in the literature for the same purpose.