The hub and casing walls of axial compressors are often modeled as smooth continuous surfaces in CFD simulations, but in real geometries, non-smooth pinches, steps and leakage cavities may exist. In the GPPS first Turbomachinery CFD Workshop, a comprehensive validation and verification campaign of RANS flow solvers was conducted, and all the simulation results consistently over-predicted the total pressure ratio at the rotor exit near the casing and the stator exit near the hub. From a recent examination of the test rig geometry, a pinched casing wall over the rotor and a leakage cavity below the stator were found, which were not considered in the workshop. In this paper, the effects of these endwall geometric uncertainties and errors are analyzed via numerical simulation. When considering the rotor casing pinch of the test geometry, the predicted total pressure ratio and choke mass flow of the compressor stage are smaller than that without the pinch, leading to better agreement with the measured data. When considering a stator hub cavity with a leakage flow rate of about 0.2% of the compressor inlet mass flow, the near-hub total pressure ratio distribution matches slightly better with the experimental data, but the effects on the global compressor stage characteristics are not visible. The relevant mechanisms of these changes in performances are analyzed in detail. The updated geometries and grids will be released to the public as a benchmark test case for turbomachinery CFD validation and verification.