Accurately modeling plasmon-modified fluorescence is important for understanding and guiding the design of experimental nanostructures that reliably enhance fluorescence. They are of particular interest due to their potential to allow localized "hot spots" of high fluorescence enhancement in a reproducible manner. Given the increasingly prevalent use of the COMSOL Multiphysics software package for simulating these phenomena, we investigate its accuracy using an analytically tractable model consisting of a gold nanosphere interacting with either a plane wave or a radiating point dipole. COMSOL simulation results were compared with a formally exact analytical theory. It was found that simulation parameters commonly used for plane-wave scattering do not necessarily produce accurate results for the nanoparticle-plasmon-coupled dipole emission case. Instead, user-input adaptive meshing parameters were found to be helpful in achieving quantitative agreements between COMSOL and analytical theory results for plasmon-modified fluorescence. Our studies suggest convergence to analytically calculated values when a minimum of two additional user-input mesh elements separate the point-dipole position and the nanoparticle surface. This practical insight is expected to aid in the application of COMSOL simulations to planning and interpreting fluorescence modification experiments.