Beam-based plasma diagnostics on ITER need high accuracy and reliability. A valid collisional-radiative (CR) model is required for ITER high energy beams. Excited states and their sublevel populations are found to be essential for some of the beam diagnostics, such MSE and CXRS/BES, and dynamics of beam penetration. Although, an assumption of statistical populations among m-states is frequently used, its validity for high beams energies and high magnetic fields is doubtful. In addition, much attention should be paid to accuracies of cross-sections used. An appropriate non-statistical nkm-resolved CR model was recently prepared [1]. Here we report an empirical verification of this model in Alcator C-Mod and the short extrapolation to the relevant parameter range for ITER. Experiment was performed on Alcator C-Mod tokamak, which operates in a unique range of parameters, well suited for testing this model for ITER beams. Beam emission spectra were collected for a selected range of plasma parameters. The ratios σ 1/ σ 0 , π 4 /π 3 and Σσ/Σπ were measured and compared to the n-resolved statistical population model and nkm-resolved non-statistical model. Measured ratios show apparent deviations from statistical model and reasonable agreement with non-statistical model.