Starting with a cube you have 8 vertices on the sphere. And 6 squares. I bet when you have subdivided them enough times, at the original 8 vertices three line segments meet, forming 120°, while at every other new vertex caused by the subdivision you have 4 line segments meeting at 90°. While the cube is a perfect polyhedron, the more complex perfect polyhedron you start with, the smaller will the difference be between the original vertices and the new ones.
When I started with an icosahedron, I had 20 triangles and 12 vertices. After each subdivision I got 4 times more triangles and... um... some more vertices. At the original vertex, 5 line segments meet, so the angle inbetween is 72° in the final (almost) sphere. At each new vertex, 6 segments meet, so the angle is 60°. I tried also with cubes - and other polyhedrons for that matter - but at least with my method the original vertices were too visible in the final image.
I bet your straight mountain range follows somewhat one edge of your original cube. And ends at one original vertex. And at the vertex you might see another line forming a 120° angle.
It might be somewhere here.
The coastline above my red line to the left would support my suspicion.
In this image of mine, that I posted earlier...
...you might find the vertices where only 5 triangles meet. They are the original ones of the original icosahedron, while at every other vertex you have 6 triangles that meet. This is a very rough model and the only thing that makes the original vertices hard to spot is the great part of flat ocean. Somewhere I have a better image with more subdivisions, but it took days to render.
And now I can't find it.