Hi Burr - this actually doesn't make a variable radius fillet - what you get with fillets on curved things like this is the same fillet radius everywhere, but some of the fillet cross-section arcs are smaller and some are larger, depending on the angle that the 2 surfaces meet at in a particular location.

Here's an illustration - here I have 2 arcs both of which have radius 5, that is they are both portions of a circle that has radius 5. But you can see that even though they have equal radius, they can have different lengths:



That's what is variable in the case you've got here - the length/size of each section (I guess you would say "subtended angle"), not the radius.

I've attached a file burr_fillet_arc.3dm which has several cross-sections of the fillet extracted as individual arc curve objects, this may help you to see more clearly what is happening exactly:



If you pick each of those arcs, you'll be able to see that each of them is of radius .125 .

And if you look from the top you'll also be able to see that they are not necessarily all oriented in a static direction like you may also have been thinking:



Hope that may help to show what is going on!

- Michael