Does that calculation take into consideration the flexing at corners or is it based upon an ideal, theoretical, unflexing shape? I would be surprised if that CAD software was doing anything more than plugging numbers into the formulas and spitting out the results. Those results don't cover everything to be considered.
I just don't think that polar moment of inertia is the whole picture. That only tells you how much the object resists. It doesn't tell what happens to the object if it is subjected to continuously varying torsion over a long period of time. It doesn't tell what happens to the steel along the edges where the planes meet after that boxed frame has been bounced down the highway a few years.
It looks to me like the greatest stress will occur at the corners. A box can flex at the corners at levels of stress far below the failure point. Isn't that what causes metal fatigue? Especially with sharp corners, isn't boxing setting the frame up to fail from little cracks, sometime down the road?