So I will just post my own learning here to help others.
What I have concluded is that for layered models, the initial smoothing operations after creating the 2D supermesh is the way to help form a good mesh.
I realize after digging that for layered meshes, "Condition Number" in the auxiliary data (Data Panel), is the better item to review when trying to create a reliable mesh for a layered mesh (compared to a completely unstructured mesh). I did do this smoothing mentioned above earlier, but found that my model was still showing a lot of spread in the condition number (from 1 to the 1000s).
Then I realized that this was a problem in only one layer. While creating the 3D mesh from the 2D mesh, I had a layer that was very thin and in some areas would "pinch" to the minimum thickness defined in the "problem settings">"Editor Settings">"Minimum slice distance" which I had set earlier to 0.1m. So when I inspected some of the larger 3D elements in my 3D mesh, I realized they were "deformed" in the z direction (they were not a nice "slab" like some of the elements in thicker layers).
So I increased my minimum thickness to 1m, re-applied my elevations, and the popup asking me how to resolve slice elevations to adhere to minimum thickness came up. I chose bottom up, and let Feflow reassign all the elevations but now with the thicker minimum thickness of 1m. So now most of the red indicating a bad (high) condition number disappeared, and the scale legend showed max values that are in the 100s instead of 1000s. Still some work to do, but much better.
In my case I have some leeway to change this minimum thickness. That may not be the case for all models. I am not sure yet if the model would have trouble converging with these high condition numbers, but hopefully I will not need to find out :)