Tompkins and Pieters (1999) identified eleven spectral classes present on the surface of the Moon. Here we model these spectral classes to determine the compositions that define them. We do this by mathematically matching spectral classes to radiative transfer computed spectra using spectral shape, contrast, and excursion parameters as defining characteristics. Model spectra are based upon known mineralogies, mafic-mineral Mg′s (molar (Mg/(Mg + Fe)) · 100), and maturities. We compare these compositional results to a compilation of ∼100 Apollo samples to determine plausible representative samples for these spectra. Results indicate that unique mineralogic solutions for several spectral classes can be determined unambiguously; however, classes previously interpreted to be pyroxene-rich are compositionally ambiguous. Results suggest these ambiguous mineral assemblages are not random, but correlate with variations in absolute reflectance. These data further suggest absolute reflectance is a compositionally diagnostic spectral parameter that should be used in conjunction with relative reflectance analysis to constrain mineralogy. Although absolute reflectance for the eleven spectral classes is not available to constrain their composition, we detail mineral, chemical (i.e., Mg′), and absolute reflectance variations of matching model relative reflectance spectra to narrow the possibilities. This information is used in a pilot study of Bullialdus crater, for which Clementine absolute reflectance is known. Results indicate Bullialdus's central peak consists of Tompkins and Pieters' spectral classes AN, AGN, and N. Bullialdus's spectra that match class N are roughly consistent with previous compositional interpretations; however, matches to classes AN and AGN indicate more mafic assemblages typical of Mg-suite norite/gabbronorite rocks.