Severe Accident Management (SAM) strategy aims to minimize exposure of public to potentially harmful radiation and the release of fission products into the environment. For assessment of SAM efficiency it is crucial to identify phenomena and parameters that present major contributions to the uncertainty in the magnitude and timing of the releases, assess their sensitivity, and quantify the uncertainty. In this work source term evaluations are performed, considering 50 uncertain input and modeling parameters that can affect the accident progression, release paths and magnitude of the release. MELCOR was used to perform evaluations for two accident scenarios initiated by: (1) large-break LOCA, and (2) station blackout, that lead to containment failure due to containment bypass or ex-vessel phenomena at RPV melt-through. Preliminary screening was performed using best-estimate and bounding assessments, where parameters were varied one-factor-at-a-time. The fraction of cesium release to the environment was found to be within ~2-35% of core inventory in case of LOCA, while it was within ~2-7% in case of SBO. This difference could be explained by the greater effect of suppression pool scrubbing in case of SBO. Releases obtained with modeling of solid core debris ejection from the vessel were marginally lower than without, for both LOCA and SBO. Dakota was used to perform Morris sensitivity analysis for the most influential parameters in the bounding assessment and the results are presented here.