Hmm. The span count on the two curves are different. The projected curve on the fuselage has a whopping 114 vertices (or so), and the other has only 22.
I am not able to change the span count on the fuselage-curve, but I can move these cv's. The trick is to keep history on and move the cv's on the projected curve (they keep themselves glued" to the fuselage surface) and not try to modify the surface itself (at least not on the outer edges, the inner cv's are modified).
um... so I have tweaked the curve cv's (proj-fuselage-cuve) and some cv's on the fillet surface itself (but only those that aren't on the edge of the surface.
Forget everything above.
The tweaking turned bad
However I identified the problem with the former fillet surface.
The problem was that maya had a "hard time" making this fillet surface AND do the tangency thing.
So I tweaked a few cv's on the end of the wing, so that any tangency to these cv's were nice and rounded and not so sharp as was the result in the former try.