An earlier modeling study of the effect of changes in macular pigment optical density (MPOD) on a wide range of surface colors is re-examined. That study reported changes in local chromaticity variance and in color spacing, some of which were incompatible with tritan-like confusions in normals associated with high-simulated MPOD. This disagreement might have arisen through the use of the von Kries correction for adaptation. The analysis is repeated, using 1782 reflectance spectra of natural and man-made colors. These colors are segregated into an array of 25 equally populated cells in an analogue of the MacLeod-Boynton cone excitation diagram. Removing the von Kries correction restores compatibility with other experimental data. Differences between the results for normal and anomalous trichromats, noted in the earlier study, are confirmed. An analysis of local chromaticity variance across color space indicates the presence of systematic patterns. The earlier study also reported differences in results across observer types (for example, between normals and protanomals) and this is addressed here by utilizing fundamentals defined by a variable photopigment template. Chromaticities are computed for the same 1782 reflectance spectra for normals and for a set of protanomals (for whom the anomalous L pigment is shifted between the normal L and M spectral locations). Colors are segregated into an array of 100 cells in an analogue of the MacLeod-Boynton cone excitation diagram. Changes in chromaticity variance with MPOD for these cells are mapped for normals and protanomals. Variance along the L/(L + M) axis is sensitive to the number of cells used for segmentation. It also increases with MPOD for normal observers but this trend reverses as the wavelength of maximum sensitivity of the L cone shifts towards shorter wavelengths (protanomalous locations).