Additive manufacturing of metals has become a key technology for advanced engineering applications, offering unprecedented design freedom and material efficiency. Despite these advantages, the laser-powder bed fusion (L-PBF) process is still affected by significant residual stresses and part distortions, whose reliable prediction remains challenging due to the highly coupled thermal and mechanical phenomena involved. To address this issue, the inherent strain method has been widely adopted as a computationally efficient alternative to fully coupled thermo-mechanical simulations. However, the predictive capability of this method is strongly governed by the calibration of the inherent strain values, as well as by the material parameters adopted in the mechanical analysis. This study presents a comprehensive assessment of the inherent strain method for the numerical prediction of distortions in L-PBF components. First, the influence of both mesh size and inherent strain magnitude on the numerical predictions is evaluated for different part geometries. The results show that some geometries are more sensitive to variations in the inherent strain values employed in the finite element simulations. Furthermore, the study highlights the mesh dependency of the calibrated inherent strain, which arises from the finite element layer thickness used to represent material deposition. Finally, a new calibration procedure for the inherent strain values is proposed, combining numerical simulations with experimental measurements of final part distortion through a fitting approach.