Propeller crashback, an emergency maneuver undertaken when a forward-moving vessel needs to stop quickly, involves reversing the propeller rotation and running the propeller in reverse into the oncoming flow. This maneuver generates low frequency, high amplitude forces on the blades which impart pitch and yaw moments on the vessel. The origins of these forces are uncertain and there are no tools based on first principles with which propeller designers can predict the maximum loadings during crashback. In order to better understand the origins of these phenomena and develop a loading prediction tool a large eddy simulation (LES) methodology using an unstructured, finite volume, incompressible LES code with a 2 nd-order accurate central difference (CD) flux reconstruction methodology, and a dynamic sub-grid scale model is being applied to crashback. In this paper, we compare results of this code with experimental data and a commercially-available unstructured finite volume LES code with upwind (UW) flux reconstruction without sub-grid scale model. Crashback simulations have been performed on the 0.3048 m diameter, zero-skew angle, 5-bladed Propeller 4381 operating at J=-0.5. We compare the mean, RMS higher order statistical moments for the integrated forces and moments to experimental data. Comparison to the experimental power spectral density (PSD) functions shows that both codes correctly predict the low-frequency blade loading and the blade rate energy concentrations. However, the CD code predicts a wider range of turbulence scales around the blade rate peak that may be important for loading dynamics. The CD LES data reveals that the flow has a bi-modal behavior that switches between vortex ring (VR) and axial jet (AJ) modes which are associated with minimum and maximum loadings, respectively.