I was trying to do some cleanup the model previously discussed at length. Here is something that is puzzling me. When I do a blendsrf, I get a straight blend rather than G1 blend.
oof… again… these surfaces are massivly massivly overdefined and the topology and isoparametric flow of your surfaces is pretty wild.
you are working way too hard. rebuild your curves. You can do this within a very close tolerance for what you are doing and build better surfaces from them. Your life will improve immensely. The end result of doing an entire model this way is going to be a model that is carrying way too much unnecessary computational overhead. Any analysis you are going to do with it down the road is going to take forever to calculate.
As part of a cleanup, cheated and added a curved/triangular extensions to the halfsiding at the top and botto of the shaft. The four 3-sided surfaces at these corner became 4-sided. Fortunately, this is an area that is a forging so there is no internal structure for this change to disrupt.