What model should I use for a 2D flow around a wing?
 

I am currently doing a 2D analysis of a race car's rear wing. Originally I performed all my tests on Fluent using the k-epsilon model. A graduate student then informed me that this model is inaccurate for flows with curved surfaces and boundary layer separation, and recommended using a Reynolds Stress Model. He said that the k-epsilon model would underestimate the drag forces. He also recommended that I should aim to get my y+ values below 1.

I ran calculations using both models and then ended up with even less drag for the RSM. The y+ values for the k-epsilon calculation were as high as 100, and for the RSM as high as 90.

My most important question is, which model is most appropriate for this flow; k-epsilon, RSM, Spalart-Almaras, or something else?

Do I need to get the y+ values below 1, and if so, how can I do this? I have been reading that sometimes this is not necessary, and these values can be within a certain range.

Note that these calculations were done at about Re=4e5, with a mesh of about 248,000 elements. I have attached images of the wing profile, the mesh, and the y+ plot from the RSM calculation.