The RELAP5-3D (version bt) computer program was used to assess a General Electric level swell experiment. The primary goal of the new assessment models was to faithfully represent the experimental facility and instrumentation. In developing the new models, a nonphysical representation of the vessel heads in a previous assessment was found. This distortion resulted in predictions that closely matched the experimental data, but were in error. The new assessment also highlighted an instability in the calculation of interfacial drag. To explore this issue, analyses were performed using three different interfacial drag correlations appropriate for large diameter pipes and/or vessels. The results of this study show that the default RELAP5-3D interfacial drag package compares most favorably with the experimental data. Additionally, a numerical instability was uncovered with the analysis performed using the Gardner correlation and was traced to the calculation of bubble diameter in the bubbly flow regime.