To create an online diagnostic system for a gas turbine, a well-formulated- influence coefficient matrix (ICM), a key component in a widely used existing theory, is essential. However, according to a recent research of the present authors, parameter deviations estimated by the traditional ICM contain unavoidable errors. These errors, called matching deviations, have been verified to be a consequence of the shifting operating point caused by component deterioration. Here we present a simple and accurate method which is still based on the widely accepted existing theory that accounts for component deterioration. Matching coefficients are introduced in this method to isolate and eliminate matching deviations. The ICM yielded by this method can improve the accuracy of both measurement prediction and parameter deviation estimations. By modeling a commercial power generation engine, a demonstration of the new method is presented and its limitation is discussed.