I just had a look at your setup. Generally, this looks as expected to me, just a couple of comments:
Using ADDBACKGROUNDGRID will re-bin also your SPH particles in the disk, so it will degrade the density field to a certain degree. The only way around this is to set up a grid yourself (i.e. just manually set the mesh generating points, but then in a volume-filling fashion).
If you don’t want to do this, another ‘problem’ is the difference in resolution between the disk and the background, which will degrade the edges. This can be improved by increasing the parameter ‘GridSize’ in the ADDBACKGROUNDGRID step. This parameter refers to the number of grid points per dimension that are added in the background, and is chosen quite aggressively with 16 in the example run. For production runs, I would recommend to choose it such that GridSize^3 is of the order of the number of grid cells in the disk (but not more, otherwise the computation would be dominated by the background).
I hope this helps. If you need further help, let me know.