Mercury has a compositionally diverse surface that was produced by different periods of igneous activity suggesting heterogeneous mantle sources. Understanding the structure of Mercury's mantle formed during the planet's magma ocean stage could help in developing a petrologic model for Mercury, and thus, understanding its dynamic history in the context of crustal petrogenesis. We present results of falling sphere viscometry experiments on late-stage Mercurian magma ocean analogue compositions conducted at the Advanced Photon Source, beamline 16-BM-B, Argonne National Laboratory. Owing to the presence of sulfur on the surface of Mercury, two compositions were tested, one with sulfur and one without. The liquids have viscosities of 0.6-3.9 (sulfur-bearing; 2.6-6.2 GPa) and 0.6-10.9 Pa center dot s (sulfur-free; 3.2-4.5 GPa) at temperatures of 1600-2000 degrees C. We present new viscosity models that enable extrapolation beyond the experimental conditions and evaluate grain growth and the potential for crystal entrainment in a cooling, convecting magma ocean. We consider scenarios with and without a graphite flotation crust, suggesting endmember outcomes for Mercury's mantle structure. With a graphite flotation crust, crystallization of the mantle would be fractional with negatively buoyant minerals sinking to form a stratified cumulate pile according to the crystallization sequence. Without a flotation crust, crystals may remain entrained in the convecting liquid during solidification, producing a homogeneous mantle. In the context of these endmember models, the surface could result from dynamical stirring or mixing of a mantle that was initially heterogeneous, or potentially from different extents of melting of a homogeneous mantle.