The stats for commits and followers (probably forks as well but since I don't have public repos so I don't know) are wrong.
For commits stat, it doesn't count the commits for an imported Git repo. Say if you import a Git repo with lots of commits already from somewhere else, it will still show 0 commit. But the followers stat will be 1 more (e.g., even if you are the only user for this repo, you will see 2 in the follower stat if imported).
For followers stat, if you cancel follow and then follow again (to your own private repo), the followers stat increases.
I believe the stats are stored separately. It is understandable for performance reason, but I think the workflow should be corrected to reflect the right stats and also you may want to do a system-wide re-calculation to fix the existing wrong stats.