The Taylor-Jakes model of two-stage melting for the generation of the mare basalts is reconsidered and expanded. Melting of the outer 1000 km of the Moon early in its history was soon followed by the deposition of a thick series of mafic adcumulates to form a Lower Mantle. Mafic orthocumulates then sank to form the Upper Mantle (60-300 km) while semi-contemporaneous crystallization of feldspar, as a mesh which did not sink and trapped minor mafics, formed the crust. The Rb-Sr model ages of about 4.6 Ga and the Eu anomalies reflect this event. The crystal fractionation gave rise to appreciable enrichment in l.i.l. elements in the pore material of the crustal and Upper Mantle orthocumulates, including a concentration of uranium that is related to the levels at which radioactive heating and second-stage melting occurred, to produce mare basalts from 3.8 to 3.1 Ga. Pore-material melting preserved the 4.6 Ga model ages. The high-Ti and low-Ti basalts are related to shallower and deeper source levels, respectively, in the mafic orthocumulate pile. Convective cells would become numerous as the thickness of the magma shell decreased, to produce complex and variable fractionating systems. Asymmetrical near-surface features such as crustal thicknesses, 'KREEP'-rock distribution, and mare-basalt eruption sites, would be expected from lateral variations in the effectiveness of cumulate fractionation at around 4.6 Ga.