This is a problem to be considered by all FEA packages. One of the reasons is because St Venant's principal does not apply near boundary conditions. The reason CREO shows high stress is likely because it is capturing the gradient better than the other packages you tried. Some things to try is reversing the constraints vs. loads, including the real boundary stiffness (mating parts), meshing this area very finely in the other FEA packages to prove the stress is really there, using non-uniform load distributions to fade out the loads. You might be able to limit the P-order and size of elements in this area to "mask" the issue but then you still cannot trust results in this area without other changes to your model. I also suggest that if there are small fillets in the ribs and intersections, you will get more realistic results to at least include those in the areas with higher stress.
Now on your design, I am pretty sure you can confirm that the displacement has been reduced with the added ribbing. (It is stiffer and appears to deflect about half as much) Keep in mind that stress "flows" and that added structure can cause stress to "flow" into the wrong areas. For more on this see this quiz. This quiz. Adding ribs and structure does not always equal lower stress. I think the diagonal ribs create some unique 3d stresses with twist whereas the middle rib is in more pure bending.
I would also consider whether you are required to have all the ribs from one side. I call this section type a "top-hat". It has better bending inertia properties than the original "C" section style you are using. Maybe you do not need the middle rib with this design.
Anyway there is my $.02 based on your question. Thanks for using the PTC community!
EDIT: Also an elasto-plastic analysis will relieve some singularity problems.
