Summary:

I CANNOT UNDERSTAD WHY BUT THE SOLUTION IS NON PHYSICAL (ALSO HAVING SOLVED THE PROBLEM WITH THE INLET).

In GoodVelocity.jpg you can find the velocity field obtained with CS 2.2.0
Using the same mesh, using CS 3 i get very different results. In the other images, you can see the main fields:

1) 	The Co Number has some problems and it's extremely similar to the parallel domain pattern (sometime i had the same problem with the old version).

2)	The velocity field is innatural. Especially notice the reattachment on the downstream side.

3)      The turbulent viscosity is quite funny around the deck. 

4) 	I initialized the flow by formula k = 1.25e-7 omega=0.004 and I used the same values for the inlet. Why do I get that gradient?

I tried to chenge turbulence model but the problem persists. 

As a general information: for the k-omega SST implemented in CS is it correct to have y+Max =5 with a one-scale wall treatement? 
I read the presentation given at Manchester on the wall treatement but I'd need some procticle advice.
Would it be a good choice to use a two-scale wall treatement for separated flows? I saw that the y+ definition changes, what y+ can i reach accordingly with the definition used with the two scales wall functions? 


