I've found the reason why the Leif's wing had a transparent underside. It was complicated.
Now, in a 3D mesh model, the faces have the so-called "normals" ('normal' is math-talk for "at ninety degrees'). These show which way the face is pointing. In many programs, such as orbiter, a face is transparent from the side opposite of the normal face.
But when I looked at the mesh in Blender, the normals looked correct. Becoming more systematic, I discovered that only some of the sub-meshes had a problem, if I manually inverted the normals, the mesh would work in Orbiter. So what was so special about these problem meshes?
Suddenly I remembered. The problem meshes had be created by the Symmetrical Modeling technique. This is a clever way to model something that is bilaterally symmetric, like an airplane or the Leif Ericson. It insures that the left and right sides are symmetrical.
It's easy. You make the crude start of the left wing. You make a "linked-duplicate" of the left wing. Then you mirror the duplicate so it becomes the right wing. From now on, any changes you make to one wing will automatically appear in the other.
So the problem was that when the *.msh export script grabbed the normals for the right wing it got a mirror image of the normals. Ick.
I'm still trying to figure out how to fix the export script.
The script still has problems with the texures and materials. If I export the mesh as a Wavefront file and use Orbiter's converter, the textures and materials look great, except it uses the first material for the entire model.
As an experiment, I exported each mesh as a separate Wavefront file, converted them, then manually combined them into one file. Too much work but the results were nice: