Hi @Anonymous
It helps to look at the definition of the von Mises stress. When using the principal stresses (S1, S2, S3), the von Mises stress is defined as follows:

Note that it uses the difference of the principal stresses, including the intermediate principal stress (S2) which you did not provide. So it is not as easy as comparing the von Mises to the principal stresses.
Secondly, you have to look at the results at the same node. The maximum von Mises stress from your results summary may occur at a different node than the maximum principal stress, which may be different than the minimum principal stress, and so on. You cannot use the maximum of each result as a comparison (unless you know the maximums are at the same node).
Thirdly, the difference between the maximum von Mises stress of 48 and 45 is not very significant in a simulation unless you have done a mesh convergence study on both analyses to assure that your results are accurate to within 10%. If you just did one mesh or were not careful to setup everything in the same way in the two analyses, you could have different results just because of the different setup. Here's another way to say this: if you give the same model to two people, the results they get could be 10% to 20% different just because they will set up the model differently (different mesh, different contact or connectors, and so on).
Your second question asked which design is more stable. How are you defining stable? The maximum displacements are almost identical (0.02739 mm versus 0.02728 mm). We have no idea where that is occurring or which direction is which, but my guess is that it is a reasonable comparison between the two analyses. Since they are "identical", I would say that each design is equally stable.
Let us know what questions you have next.
John Holtz, P.E. Global Product Support
Autodesk, Inc. If not provided, indicate the version of Inventor Nastran you are using.If the issue is related to a model, attach the model! See What files to provide when the model is needed.