For triangulation you need the normal for each point of the 3D surface.
The original algorithm for implicit surfaces is based on this.
Addon for triangulation of implicit surfaces/ forms with holes
For sphere and cylinder the normal is very easy to determine.
But it would also be enough to determine the normal somehow approximate.
But also note the documented errors if the outline is too jagged. In your example I see sharp points, much more problematic than in my star example.
