The compositions of plagioclase, the most abundant mineral in the Martian crust, reflect changing conditions during magmatic evolution. Plagioclase contains spectral features at thermal infrared wavelengths that permit its detection by thermal emission spectrometers (Thermal Emission Spectrometer (TES), Thermal Emission Imaging System (THEMIS), and Miniature TES (Mini‐TES)) on Mars spacecraft. Previous studies have determined the accuracy with which average plagioclase compositions can be modeled in simple two‐component sand mixtures and terrestrial volcanic rocks. Studies of terrestrial rock analogs suffer from difficulties in accurately determining the average plagioclase composition for comparison with the spectrally modeled composition. Sand mixtures, however, provide a means of controlling plagioclase compositions for direct comparison to those modeled by linear deconvolution. This has allowed us to address how compositional complexity may affect our ability to derive average plagioclase compositions from thermal emission data. In this study, we examine the accuracy with which average plagioclase compositions can be modeled from emission spectra of complex mixtures of three, four, and five compositions of coarse (500–850 μm) plagioclase sand. Additionally, we examine multiphase mixtures of plagioclase with pyroxene, olivine, magnetite, and ilmenite that are analogous to selected Martian surface materials. Increasing the number of plagioclase compositions or introducing additional mineral phases does not affect the accuracy previously reported for modeling average plagioclase compositions. Plagioclase can be modeled to within 6 An of measured compositions at laboratory, TES, THEMIS, and Mini‐TES resolution (within 1σ standard deviation).