Bad elements along a curved vanes
 

I have created mesh for a casing treatment. After I create prism layers, minimum angles become too low, which I believe it may cause convergence problems.
Bad elements exist along the curved vanes inside the casing treatment.
I have tested curvature refinement, increased the number of elements in gap but still bad elements exist there. I have tried many options but still I am unable to find the cause.

How to improve mesh quality?
 

I followed Simon's presentation of no pyramids and still my mesh quality has not improved.
I have used octree, delaunay, advancing front mesh algorithms, curvature proximity based refinement, auto reduction but still there are some irregular grids as shown in the attached photos.
Does anybody have any idea why this occurs?

I know similar questions to this may have been asked many times on the forum, but after many trials I have not been able to figure out the cause of the problem.
I am using CFX to perform a simulation on a type of casing treatment with vanes.
I have done many trials but elements with minimum angle of 0.25 still exist in the lower end of the vanes which I think will cause problem with CFX.
I have done:
smaller tetra elements on the surface of the vanes
increased the number of nods along the curve intersecting vane surfaces
defined thin cuts with the two vane surfaces in different parts
defined curvature/proximity refinement with 5 elements in gap
It seems that just the prism layers have minimum angles of 10 degrees and the bad angles come from tetra and pyramids. So, I followed Simon’s presentation on reducing pyramid elements to reduce pyramid elements but nothing has changed.
Can someone please suggest any solution?
The first and second photos are with the prism layers activated on the vane surfaces.
The third photo is without prism layers. As can be seen few elements have less than 18 degrees in this case.