Loss-of-Coolant Accident (LOCA) phenomena and the system’s response are critical for the prediction of pressurized thermal shock (PTS) in nuclear power plants. PTS can lead to catastrophic scenarios, especially in aging plants. To address this, numerous experiments have been conducted over the years at dedicated test facilities to develop codes capable of predicting the progression of these accidents. This work analyzes four tests conducted at the UPTF (Upper Plenum Test Facility)[1] thermal - hydraulic experimental facility which aimed at reproducing the low pressure end-of-blowdown, refill and reflood phases of a LOCA. The first part of the study details the facility setup, experimental procedures and available data. Additionally, a physical description of the key two- phase phenomena is provided. Subsequently, the study focuses on the facility modeling with the French system code CATHARE. Emphasis is placed on translating the model to the latest code version and improving its accuracy and predictive capabilities. The improvements are made to refine loop meshes, pressure drops, boundary conditions and injection laws. The CATHARE results are compared with the experimental data and one-at-a-time (OAT) sensitivity analyses are performed. In particular, a more in-depth analysis is conducted on the pressure drops, the condensation models and the stratification model. The impacts of the new CATHARE condensation model IPTSCOND2021 and the stratification model STRATREV63 are evaluated. The final part of the study focuses on the uncertainty propagation and sensitivity analysis using the GRS (Gesellschaft für Anlagen- und Reaktorsicherheit) method and the Wilks theorem [2]. The influence of the key parameters is assessed using the Spearman and Pearson correlation coefficients. The study confirms significant improvements in code’s performance, particularly in calculation time. The findings highlight that the pressure drops and the condensation model are the most influential parameters during the studied LOCA scenarios.