Let L be the set of all software for Linux and W be the set of all software for Windows.
If there is a subset of L that is larger than W, then L itself is also larger than W.
I was proposing that if you count every image, or every repo, then dockerhub-only or the GitHub-only subsets of L might be larger than W because the effort of creating an additional windows project is much greater than the effort of creating an additional Linux project.