Excitation-emission matrix (EEM) fluorescence spectroscopy coupled with parallel factor (PARAFAC) analysis has been widely applied to characterize dissolved organic matter (DOM) in aquatic and terrestrial systems. However, its application in environmental samples is limited because PARAFAC is not able to handle nontrilinear EEM data, leading to the overestimated number of components and incorrect decomposition results. In this work, a new method, parallel factor framework-clustering analysis (PFFCA), is proposed to resolve this problem. First, simulated data with different signal-to-noise ratios and intensities of nontrilinear structure were tested to confirm the robustness of PFFCA. The residual sum of squares (RSS) of PARAFAC was significantly higher than that of PFFCA (p < 0.037). Second, a set of data originating from a synthetic mixture of humic acid and bovine serum albumin was applied to compare with PARAFAC with known samples. PFFCA provided an estimation (R2 > 0.92) closer to actual EEM than PARAFAC (R2 > 0.81). Finally, to confirm the feasibility of PFFCA in analyzing natural samples, DOM-containing samples collected from both a polluted lake and river were tested, indicating that PFFCA provides a more precise estimation than PARAFAC. The results clearly indicate that PFFCA offers a robust approach for the unique decomposition of complex synthetic and natural samples, which is of great significance in understanding the characteristics of DOM in aqueous systems.