This is not an attack. I see the fillet improvements in V6. I just think that FilletEdge is not a good fit with surface modeling. I’ve heard people complain about how much better fillets are with expensive solid modeling apps, but I’ve taken my NURBs into those apps, and the fillets fail there as well.
Speaking for jewelry designers, even the good modelers create nasty corner conditions. All of us start modeling in Rhino using some plugin’s automated workflow that creates edges which are best described as a bowl of bananas. The error rate of fillets will always be high enough that we stop attempting to use it.
Instead of asking for fillets to be better, I accept the complexity of the problem and ask McNeel to come up with something entirely different. Maybe something quick and dirty, like a bulletproof way to knock off a corner of a closed polysurf without perfect G2 continuity at an exact radius. Or look at all the ways people solve this problem manually and creating a workflow that speeds up the tedious nature of the process.