Lets start with your geometry, of the three geometries, I would not recommend using “AssembledCar_Delta3” and “AssembledCar_delta_ln3” due to the former being not watertight and the latter as a STL file which is notoriously difficult to work with.
Now onto your mesh. Observing flow separation and vortices usually require a few key things to consider like appropriate selection of the turbulence model, the approach taken to resolve the viscous sublayer near the car, mesh sensitivity, convergence and selection of transient or steady-state simulations all of which adversely affect your results significantly.
Simply increasing refinement areas as you have done in your meshes will yield good results as flow approaches the vehicle but forgoing things like the boundary layer inflation and its ties to the y+ value based on the turbulence model selected will heavily affect your results and will likely not produced accurate enough results.
With regards to your simulation, I see you have selected K-omega SST which is a relatively good turbulence model. However you do very much need your boundary layer inflation and your corresponding y+ value in order to produce accurate results. In your simulations as well, why it is stopping is due to your maximum run time being too low, adjusting that to say 80,000 fix this issue and allow your simulation to be complete.
These come directly from literature or journals done by people who have done such simulations. The equations are specific for different cases so you’ll have to search for them.
The last thing you need to take note is whether your vehicle is supposed to be locked down on the floor or simulating moving. If the latter is needed then you will need wheels and simulate that as well as a moving floor to simulate your car moving.
Do calculate and perform your boundary layer inflation to a acceptable y+ value (typically 30 to 200 for a wall function approach) and adjust your maximum run time to allow your simulation to complete.