An experimental campaign on a linear high-pressure turbine cascade showed that downstream flow is affected by pressure distortion, non-periodicity, and shock–wake interactions, which significantly influence aerodynamic losses. To complement these findings, steady unstructured Reynolds-Averaged Navier–Stokes (RANS) simulations were conducted with two goals: to compare mesh refinement strategies and to evaluate loss variations between periodic and non-periodic configurations. For the periodic case, two refinement methods were assessed: Bodies of Influence (BOI) and Adaptive Mesh Refinement (AMR). Both ensured mesh convergence and accurate capture of flow structures, though BOI required manual placement of refinement zones, while AMR relied on suitable detection criteria. AMR was ultimately preferred for its broader applicability to complex shock systems.