As @HughesTooling and @TrippyLighting report - the surface normals are normally automatically corrected when stitched to a solid.
In the early days of Fusion - I used to always manually correct before stitching.
In past couple of years I have only seen issues when the bounding edges were ambiguous.
Remember - the computer cannot see. It is only following some algorithm.
What is obvious to us visually might have no logical meaning to the algorithm.
Perhaps @jeff_strater can offer to take a non-disclosure look at the geometry.
Diagnosing issues like this helps improve the algorithm.
Edit: For others who find this thread via a Search...

a Surface has a front side and a backside.
After stitching into a solid - we no longer need to see the backside of the solid body Face, so to speed things up there is no need to shade (appearance) the backs of the faces. But if the face normal is for some reason incorrect we cannot see it because we are looking at the unshaded back face. There are a couple of ways to test if this is the issue involving steps to get a look at the incorrect front side of the face. But I can envision (pun intended) if the geometry is corrupted - the simple test might not work.
When using Reverse Normal - note the different colors on the same surface for front side and back side.


If I remember correctly - in early versions of Fusion it used to display this front side/back side colors without having to go into Reverse Normal. To me this is a regression in functionality.
In the other CAD programs that I use - the Workplanes also have a front side and a back side that makes flipping entire sketches easy.