Using curvature combs like that is cool. With that shorter shoulder loft, I forgot to use a centerline rail as I did before. So I tried it again, this time making a rail via a spline from the intersection curve of the sweep, to the top loft intersection line. I tried two different continuity constraints, first tangent for the spline, then smooth. The latter gave better results.
So here is an off center curvature test for the rail spline with only tangency constraint that is rough:
And here is the result with smooth spline end constraints.
In order to adjust the weight of the smooth coincident constraints, I followed the procedure Nathan showed at:
https://screencast.autodesk.com/Main/Details/c4084682-ca66-4b0c-890c-0bb0a5107dda
adjusting the weight handles until got a good curvature (and being sure to first fix/anchor the curves and curve endpoints).
This was the curve where part was used as a centerline rail.
I'd say the results are pretty good. I would be interested if need to even more close to perfect, if additional rails from splines also with smooth end constraints would lead toward that goal.
Jesse