This study presents a comprehensive validation of the RELAP5/MOD3.3 system thermal-hydraulic code against the ATLAS Test C2.3, which simulates a Double-Ended Guillotine Break (DEGB) in the main steam line of an APR1400 steam generator with Passive Auxiliary Feedwater System (PAFS) activation. The validation effort emphasized refining the input system model to more accurately reflect experimental conditions, highlighting RELAP5's capability to simulate complex passive safety scenarios with high fidelity. A pre-test simulation was first performed using nominal input data; while it reproduced general system trends, it exhibited notable deviations in steam generator water level, pressure response, and passive flow behavior. A post-test model was then developed with improvements including an updated decay heat profile, corrected heat transfer areas, enhanced control logic for PAFS and MSSV actuation and break modelling. The post-test simulation showed excellent agreement with experimental data, maintaining an error margin within +/- 2.5 % across key thermal-hydraulic parameters such as pressure, temperature, flow rate, and water level. In addition, a code-to-code comparison with TRACE was conducted using identical boundary and initial conditions. The RELAP5 simulation with the refined system model demonstrated comparable or improved accuracy in capturing critical phenomena, including asymmetric depressurization and PAFS-induced natural circulation. This work underscores the importance of detailed input modeling in achieving high-fidelity simulations and confirms the suitability of RELAP5/MOD3.3 for evaluating passive safety systems in advanced reactor designs.