Multiple end-member spectral mixture analysis (MESMA) was applied to the Clementine UVVIS global 1 km multispectral data set, and the resulting highland material fraction image was used to investigate highland contamination of mare surfaces by impact cratering. MESMA decomposes each pixel with the number of end-members fewer than the number of the spectral bands of Clementine UVVIS data. This allows the use of variable end-member combinations to accommodate global spectral variance. A 13-end-member set of lunar soil components that spans the lunar spectral diversity was selected from the UVVIS data and included 1 mature and 4 fresh highland soils, 4 mature and 2 fresh mare soils, and 2 soils of dark mantling materials. This set was applied to the lunar data through MESMA, and an aggregated global highland abundance was derived. Comparison of the highland fraction results from MESMA and traditional spectral mixture analysis (SMA) demonstrated that MESMA better accommodates the compositional variation of mare soils, resulting in a greater accuracy in the measurement of highland. With the derived products, we investigated highland contamination in Mare Nectaris, Mare Fecunditatis, and Mare Crisium. These mare surfaces are proximal to the large craters Theophilus, Taruntius, Langrenus, and Proclus and provided an opportunity to investigate highland contamination by large impacts. The analyses indicated that Theophilus impact resulted in highland contamination of 20--80% on most of Mare Nectaris, while Taruntius and Langrenus impacts caused 5--40% contamination of Mare Fecunditatis. The trend of minimal highland abundance in Mare Fecunditatis correlates poorly with basalt thickness, suggesting an efficient lateral mixing due to the large craters Taruntius and Langrenus. Additionally, Proclus, combining with other craters, was responsible for highland contamination in most of the Mare Crisium surface, and some basalt classification in this mare reflects highland contamination of the spectral properties. MESMA significantly reduces uncertainty in calculating global composition variations due to mixing, and these analyses demonstrate that large impacts played a dominant role in delivering highland materials onto three mare surfaces.