I've used RISA 3D for years and have a high trust in its performance and accuracy.  I would bet you have a modeling issue.  One of the best things to do is to always first look at your models deflection as this usually reveals how the model is being interpreted and how the matrix solution is panning out.

Per your statement "pushing everything to the corners" ...  remember also that force will follow stiffness.  So the stiffest load path to the boundary conditions will take the most load.

I've also been using RISA for 10+ years and every funky result was caused by a misfit model. There was one time that my RISA output of a truss with unusual geometry didn't at all match someone else's RamAdvanse truss model (the RamAdvanse model was giving us 0 force web members) but I still trusted the RISA results.  

If you have a large shear due to wind and just the end bays braced that's where the force will go.  

Thanks for your responses.  Please don't get me wrong, I have been a fan of RISA for quite some time (if anyone remembers when RISA2D was new software).  I go back to the time of slide rules, Smoley's Tables, and Roark's Formulas for Stress and Stain.  In a 2D world you would design the four sided rigid frame to take the winds and dead loads on the four columns supporting the stack.  If you use RISA plate improperly it will give you a bogus answer.  

I also believe GIGO may be the problem with the model.  

I agree with JAE.  The plate walls may have not been sub-meshed or stiffened enough to keep down the deflection in the model.  FYI, The model in a deflected shape looks like a sheet in the wind attached at four corners.  Links from the columns or stiffeners to the rectangular wall plates are on the four corners of the plates.  If you impose wind loads on the plate in a 150 mph wind zone it is going to find a way to the base plates (even if it is the four corners at the end walls).

I am at a slight disadvantage on evaluating the model in question.  I do not have RIS3D on my computer at the office, and I have been spoon-fed print-outs or looking over someone's shoulder.
   
I was planning to push for a conference call with RISA Technical next week.  But, I have received so many good responses from Eng-Tips in the past that I thought I would give it a try.

Thanks again  

It's sort of difficult to envision your model.  If you can post it on a web-photo sharing site (www.villagephotos.com) and link the URL maybe we could take a look at it and get a better view of what you have.

Are you making a finite element model with plates?  I have only used that part of RISA once, and that was to analyze a glulam beam that had a huge hole cut in the middle.  I bet I put 40 hours into that model.  Before I cut the hole in the middle, I analyzed the beam by hand and with RISA (the typical matrix analysis part of it) and messed around with my plate model until I got results that essentially matched the other, then I cut a hole in it.  

Unless you really know what you are doing, I think it would really be good idea to first get your model to give you the results you already know and then modify it.  Also read the help for using plates and make sure you aren't doing anything they tell you not to, like use them for diaphragms.  Also, the smaller the mesh the better and no triangles.