Hi,
Finally I got the simulation running with an uploaded mesh! 
Most of my problems came out due to a very tiny feature that was created in the CAD design which led to some tets with an extremely high aspect ratio (over 200 and sometines even 1000). After cleaning the CAD the mesh came out with very nice aspect ratios (below 4). The mesh importation process worked without any trouble and the simulation run smoothly with very nice results!
I think that in this case, where I wanted to compute the contact stresses between an aspherical glass lens and a conical metal surface, the use of a structured mesh in the contact interfaces is justified since using the unstructured meshes given by the standard mesher yields an irregular contact area with some points overstressed. This is not what should be expected according to the Hertz theory, considering this is a case of cylindrical contact with a line of contact.
Below some results are presented of the 1st principal stress in the lens. Both have the same load applied (-55N in a 12º wedge simplified geometry). Penalty method is used with a Penalty Coefficient of 1e13.
Unstructured mesh
Structured mesh
The results expected according to Hertz contact theory, however, were a 1st principal stress of 5.62 MPa instead of 2.9 MPa. I tried changing the value of the Penalty Coefficient and it has a massive effect on the results! A value of 1e14 yields a principal stress of 5.67 MPa (the expected value according to the Hertz formulation) and a value of 5e14 gives a principal stress of 8.59 MPa…
This disparity makes me wonder what value for the Penalty Coefficient should be used and if theare is a way of estimating the proper value. I know that the documentation gives us a starting point of 5-50 times the Young Modulus, but I would like to know something more accurate…
Besides, I don’t know if the issue with the principal stress vectors comented here has any effect on the results. I hope it gets solved as soon as possible…
Again, @ggiraldof thanks for your support!
Best,
Alex