An analysis of an anomaly in the subcooled critical flow model in the RELAP5/MOD3 computer code is presented. Specifically, the code produces a discontinuity in going from unchoked subcooled liquid flow (i.e., subsonic flow) to subcooled choked flow (i.e., sonic flow). The same anomaly has been reported elsewhere. The root cause for this anomaly has been analyzed, and it is found that the user-supplied junction loss coefficient and discharge coefficient play an important role in the occurrence of this anomaly. The analysis is verified by assessment against a test problem simulating single-phase liquid flow through a convergent nozzle with a fixed upstream pressure and a varying downstream pressure. A corrective measure to eliminate the discontinuity is suggested.