Internal consistency of thermodynamic data has long been considered vital for confident calculations of aqueous geochemical processes. However, an internally consistent mineral thermodynamic data set is not necessarily consistent with calculations of aqueous species thermodynamic properties due, potentially, to improper or inconsistent constraints used in the derivation process. In this study, we attempt to accommodate the need for a mineral thermodynamic data set that is internally consistent with respect to aqueous species thermodynamic properties by adapting the least squares optimization methods of Powell and Holland (1985). This adapted method allows for both the derivation of mineral thermodynamic properties from fluid chemistry measurements of solutions in equilibrium with mineral assemblages, as well as estimates of the uncertainty on the derived results. Using a large number of phase equilibria, solubility, and calorimetric measurements, we have developed a thermodynamic data set of 12 key aluminum-bearing mineral phases. These data are derived to be consistent with Na+ and K+ speciation data presented by Shock and Helgeson (1988), H4SiO4(aq) data presented by Stefánsson (2001), and the Al speciation data set presented by Tagirov and Schott (2001). Many of the constraining phase equilibrium measurements are exactly the same as those used to develop other thermodynamic data, yet our derived values tend to be quite different than some of the others' due to our choices of reference data. The differing values of mineral thermodynamic properties have implications for calculations of Al mineral solubilities; specifically, kaolinite solubilities calculated with the developed data set are as much as 6.75 times lower and 73% greater than those calculated with Helgeson et al. (1978) and Holland and Powell (2011) data, respectively. Where possible, calculations and experimental data are compared at low T, and the disagreement between the two sources reiterates the common assertion that low-T measurements of phase equilibria and mineral solubilities in the aluminum system rarely represent equilibrium between water and well-crystallized, aluminum-bearing minerals. As an ancillary benefit of the derived data, we show that it may be combined with high precision measurements of aqueous complex association constants to derive neutral species activity coefficients in supercritical fluids. Although this contribution is specific to the aluminum system, the methods and concepts developed here can help to improve the calculation of water-rock interactions in a broad range of earth systems.