Cascade testing and flow analysis of three high-turning supercritical compressor blades were conducted. The blades were designed at an inlet Mach number (M1) of 0.87 and inlet flow angle of 48.4 deg, with high camber angles of about 55 deg. The baseline blade was a conventional Controlled Diffusion Airfoil (CDA) design and the other two were optimized blades. The blades were tested for an inlet Mach number range from 0.61 to 0.95 and an inlet flow angle range from 44.4 deg to 50.4 deg, at high Reynolds numbers (1.2-1.9x10^6 based on the blade chord). The test results have shown lower losses and better incidence robustness for the optimized blades at higher supercritical flow conditions (M1>0.83). At the design condition, 30% loss reduction was achieved. The blade-to-blade flow was computed by solving the two-dimensional steady Navier-Stokes equations. Experimental results, in conjunction with the CFD flowfield characterization, revealed the loss reduction mechanism: severe boundary layer separation occurred on the suction surface of the baseline blade while no separation occurred for the optimized blades. Furthermore, whether the boundary layer was separated or not was found due to different shock patterns, different shock-boundary layer interactions and different pressure distributions on the blades. For the baseline blade, the strong passage shock coincided with the adverse pressure gradient due to the high blade front camber at 20% chord, leading to the flow separation. For the optimized blades, the high blade camber shifted to more downstream (30-40% chord), resulting in stronger flow leading edge acceleration, less strength of the passage shock near the blade surface, favorable pressure gradient right after the passage shock, thus no flow separation occurred. The flow understanding obtained by the current research can be used to guide the design of high-turning compressor blades at higher supercritical flow conditions.