Tests were performed in a non-axisymmetric, single nozzle, strut-based ejector to investigate mass flow entrainment, choking mechanisms and stream mixing as a function of primary (strut nozzle) to secondary (duct inlet) flow stagnation pressure ratio. Experimental results show a mass flow choke in the mixing duct rather than a traditional aerodynamic choke in the strut gap. The stream mixing length was constant for lower primary flow pressures, whereas mixing length varied with pressure at higher values. A companion numerical study was performed using Reynolds Averaged Navier-Stokes solutions to investigate several turbulence models. Based on both 2-D and 3-D simulation results, compressibility correction to conventional incompressible twoequation models was required for capturing the supersonic ejector mixing phenomena. The Baldwin-Lomax and the SST two-equation models were capable of capturing the essential flow features. Even with compressibility correction, the k-Σ model could not reproduce wall-dominated phenomena such as mixing duct pressure recovery.