The measured loss characteristic in a high-speed cascade tunnel of two turbine blades of different designs showed distinctly different trends with exit Mach number ranging from 0.8 to 1.4. Assessments using steady Reynolds-averaged Navier--Stokes equations (RANS) computation of the flow in the two turbine blades, complemented with control volume analyses and loss modeling, elucidate why the measured loss characteristic looks the way it is. The loss model categorizes the total loss in terms of boundary layer loss, trailing edge (TE) loss, and shock loss; it yields results in good agreement with the experimental data as well as steady RANS computed results. Thus, RANS is an adequate tool for determining the loss variations with exit isentropic Mach number and the loss model serves as an effective tool to interpret both the computational and the experimental data. The measured loss plateau in blade 1 for exit Mach number of 1–1.4 is due to a balance between a decrease of blade surface boundary layer loss and an increase in the attendant shock loss with Mach number; this plateau is absent in blade 2 due to a greater rate in shock loss increase than the corresponding decrease in boundary layer loss. For exit Mach number from 0.85 to 1, the higher loss associated with shock system in blade 1 is due to the larger divergent angle downstream of the throat than that in blade 2. However, when exit Mach number is between 1.00 and 1.30, blade 2 has higher shock loss. For exit Mach number above an approximate value of 1.4, the shock loss for the two blades is similar as the flow downstream of the throat is completely supersonic. In the transonic to supersonic flow regime, the turbine design can be tailored to yield a shock pattern the loss of which can be mitigated in near equal amount of that from the boundary layer with increasing exit Mach number, hence yielding a loss plateau in transonic-supersonic regime.