A spectrum filtering technique has been introduced to improve harmonic analysis in this study. The impact of site characterization and physical processes have been examined to justify the necessity of applying the spectrum filtering technique following classic harmonic analysis. Investigations show that the imperfection of the harmonic toolbox, e.g. T_Tide, could be attributed to significant contributions to the long period energy of density gradient and background currents, the greater variability of current amplitude energy, and the larger energy surrounding the clustering diurnal and semi-diurnal tidal constituents. From applications on hydrodynamic model data, this study supports the argument that harmonic analysis with the spectrum filtering technique could significantly improve the accuracy of tidal energy and residuals separation.