circleinasquare
I have been watching your post and have some comments / suggestions hopefully to help vs confuse:
More segments are not better. As that number increases the size of model MB wise gets larger, you run into problems of end point inference ( even to point of snapping to undesired point, color comes in handy here , some printers limit the number of triangles in the model ( ~ 1,000,00), and other issues like Box noted above plus many more like your graphics card performance. While you get up to speed keep segment count down to default;

The issue of axis color is some what misleading. A color axis does not mean it is " on axis" per se but is parallel to that axis and so one can draw a infinite number of them. It is a another tool in the Su tool box to help along with the other inference tools. Just draw your profile on a plane and make sure it forms a face. It may even help to use small ref rectangle to start, the arc tool can do its thing some times and that will give you a on face ( plane ) inference point;

There are no tangents in SU in the classical math sense we were taught in school ,but that face profile must be perpendicular to the edge you are extruding around to start ( so if you attempt to start at a section of the segment line you have to make sure it is perpendicular to it. The first step in that extrusion process is a projection of the face along the first vector and will get changed in size if it is not or may not even work.

Instead of drawing the profile perfect to start( like you are doing with large seg count=> inference problem) start with a straight line approximation then lay in you arcs, circles etc.

Once you have a good SU model run it through the link I gave you above and it will give other info on printing acceptability( You can make a SU solid model that Printer's will not accept). That's one of many and is free.
Just some thoughts