I went back and tried fixing that model shown in the screencast. I've tried before but didn't have success. I think I found the problem. Basically the model with the airfoils projected onto cylinders will only allow a root fillet radius of .08mm while the model with the airfoils not projected onto cylinders will allow a radius of 1mm. You can actually see where the fillet breaks in some cases. I just kept trying different fillet values. Basically, I tried a lot more values than I did in the past. The thing is, .08mm is really small and undesirable. That's why I never went that low before. This was all done with Rhino 5, but in the past it seemed to work almost exactly like MoI v2 trial. I've been waiting for MoI v4 to come out, so I can switch from Rhino to MoI.

I don't know if MoI behaves the same as Rhino wrt fillets. Rhino will create invalid fillets with no warning. Calculating volume is how I check to see if the operation was actually valid. As mentioned before, with Rhino 5, I have had some models calculate a volume but still have an invalid step file exported. So this whole root fillet thing is a giant pain in the butt.

But the main thing is that you have to experiment with root fillet radius values. You also get limited by the te radius. So it's a lot of trial and error, even if you created good geometry up to this operation. One other thing I've found as far as model quality. When you go to fillet that edge, it should be one curve when you select it. If the model has you selecting multiple edges, chances are it's not going to fillet correctly. I also found the blend edge was no different than the fillet edge.