The Southern Variant of the Forest Vegetation Simulator (FVS-SN) is made up of individual submodels that predict tree growth, recruitment and mortality. Forest managers on Ft. Bragg, North Carolina, discovered biologically unrealistic longleaf pine (Pinus palustris) size-density predictions at large diameters when using FVS-SN to project red-cockaded woodpecker (Picoides borealis) habitat. Inventory data from Ft. Bragg indicated the mortality submodel was responsible for the over-predictions. Three approaches to remedy longleaf pine mortality predictions in FVS-SN were explored: (1) using stand density modifier keywords, (2) using a tree size cap to influence mortality rates but not growth, and (3) iteratively invoking a mortality rate based on empirical data. Results showed the third approach was the only viable alternative. Details of this approach are described so that an FVS-SN user can effectively constrain predicted longleaf pine size-density combinations at realistic levels. Although the approach was successful, it required advanced knowledge of size-density relationships for longleaf pine. It also demands an advanced understanding of FVS-SN from the user. We suggest over-prediction of size-density relations at large diameters will be evident in any growth and yield model using similar mortality logic. Therefore our results provide a general framework for improving the accuracy of mortality predictions in FVS.