Thanks, Rob. Actually, I'm doing all of this because I'm trying to simulate a thermal plume above a human model in a room. It is a natural convection case and I'm having huge problems in order to achieve a converged steady-state solution. I talked with my colleagues and they suggested making the mesh perfect in order to isolate the problem (they think that I'm using the wrong turbulence models or solver settings). So I tried to make a "perfect" mesh so I can find more easily where my problem is.
But that is another topic, in this forum question I asked about mesh generation so I won't bother anymore with my CFD simulation.