Lid driven cavityflow problem
 

I'm trying to solve the lid driven cavityflow problem using the projection method. The space derivatives have been discretized with a meshless least square based finite difference scheme. I have applied the vorticity streamfunction method, which worked very well. However, when I try to solve for the primitive variables with the projection method, the results gets weird. The solution is somehow almost symmetric about the vertical centerline, no matter what viscosity I use. Smaller viscosity gives a "squized" version of a solution with higher viscosity. Is there anybody who might have an idea of what this could be caused by?

Regards Paal