When performing numerical simulations of the aerodynamic properties of vehicles, the simulation domain used is often a large box with a very low blockage ratio and a fully moving ground plane, replicating open road conditions. However, the physical measurements to which the simulation results are usually compared are typically performed in wind tunnels, with deficiencies concerning blockage, ground modeling, and other boundary interference effects. Some of these effects can be corrected for, but such corrections are usually performed on a global level and thus fail to correct for local effects that might influence different configurations of the vehicle in different ways. In this work, the typical open road numerical setup is compared to simulations where the computational domain is a virtual model of the complete slotted wall wind tunnel test section geometry. A vehicle of sedan type is simulated in different configurations, and the simulation results are compared to forces and pressure measurements from physical tests. The results show that the absolute drag coefficient can be predicted with very good accuracy by simulating the car inside the wind tunnel if compared to uncorrected measurement data. However, despite the good agreement for drag, the prediction of lift is not as satisfactory.