BOB Z MIGHT HAVE TO LEARN TO USE THE ATOMIC BOMB OF FILLETS. WORKS EVERY TIME.
EXTRUDE A BLOCK OF MATERIAL UP OUT OF THE SURFACE. THE BLOCK IS A STOP FOR THE FILLET. IT MUST BE BIG ENOUGH TO STOP THE .25 FILLET. PLACE THE BLOCK IN THE LOCATION WHERE THE FILLETS HAVE TROUBLE.
FILLET UP TO THE BLOCK FROM ALL DIRECTIONS.
THEN USING THE EDGES OF THE FILLET WHERE THEY INTERSECT THE BLOCK AND ANY OTHER EDGES NECESSARY CREATE A FILL SURFACE FEATURE.
USE THE FILL SURFACE FEATURE TO TRIM AWAY THE PROTRUDING PART OF THE BLOCK.
P.S. This was something Ed Eaton showed me. I don't take credit; I just use it.
PPS If fillets as large as you need were formed there, would any of the starting surfaces be consumed and changed in form? In other words, does doing what you are going to do change the design, form fit and function?
PPPS There are a lot of little sliver surfaces in your imported part. These are not good. Prior to any work did you run a TOOLS/CHECK with Verification on Rebuild turned on? CTRL Q. Import diagnostics. Did you try exporting to VDAFS? VDAFS has a very stringent checker. It may turn up trouble that other means won't. Did you turn on surface curvature shading and look for "whopper jawed" surfaces? This is just GIGO checking.
PPPPS You can always load a trial version of Rhino and fix it. Sometimes I wonder about jobs like this. There is really no design intent issue and history based modeling can get in the way more than help. More than that, this kind of imported geometry more often than not is not conducive to SW way of doing things for many reasons.