Aerodynamic stall is a leading cause of fatal accidents in aviation. This has introduced regulatory changes to pilot training, where aerodynamic stall models are utilized in full flight simulators for mandatory stall recognition and recovery training. In this research, two longitudinal aerodynamic stall models of the Cessna Citation II were compared side by side to assess the fidelity of two different flow separation modeling options. The first model contains one flow separation state for the entire wing, while the second model models wing and stall-strip effects as two separate states. In previous research, the second model type showed approximately 30% improvement in matching flight test data. Pilot-in-the-loop tests were conducted in the SIMONA Research Simulator, where a Cessna Citation II test pilot actively flew quasi-steady stalls with both models. The tests revealed that the two-state flow separation model had better stall onset characteristics, but it lacked in recovery performance compared to the one-flow-separation model, often leading to higher oscillations. In addition, the tests showed that for an aircraft with reversible controls, an integrated control column with control loading and buffet feedback is required to further improve the model fidelity. The test data analysis revealed that elevator effectiveness variation, buffet cues, and the flow separation hysteresis time constant were the main contributors to oscillations in stall onset and recovery characteristic of the two-state flow separation model. However, the direct effects of having two flow separation states instead of one could not be fully isolated due to pitch model difference between the two stall models. The findings of this research provide guidance for future stall model development aimed at improving the fidelity and training effectiveness of full flight simulators.