On-orbit calibration techniques for star trackers are critical for precise attitude determination in artificial satellites. As calibration algorithms based on interstar angles do not require a priori information about the satellite attitude, they are a focal point for studies in this area. In these algorithms, the distance between each ideal star point and the principal point on the focal plane is required to model lens distortions. These distances are typically approximated using the distances between observed star points and the principal point, named the simplified approximation method. We propose an embedded iteration method and perform numerical simulations and ground-based night-sky experiments to compare the results of star tracker calibrations based on the two proposed methods. The results indicate that the simplified approximation calibration model has systematic model errors, which are predominantly manifested in the principal point estimates. Nevertheless, the simplified approximation converges more rapidly and easily than embedded iteration. Therefore, embedded iteration should only be used if the calibration accuracy is critical, large lens distortions are present, and ample computational resources are available. Otherwise, simplified approximation should be used to model lens distortions.