I combatted this in two ways - first, I only looked at the top 433 reddits.
There are always going to be the same people cross-subscribed between A and B as between B and A. This graph is not of the number of people cross-subscribed between two reddits - it's of the sum (number of people cross-subscribed)/(users in A) + (number of people cross-subscribed)/(users in B). So if a lot of people in a tiny subreddit are cross-subscribed, they get a big boost from the first term, but almost no boost if they make up a tiny sliver of subscribers to reddit B.
There are always going to be the same people cross-subscribed between A and B as between B and A. This graph is not of the number of people cross-subscribed between two reddits - it's of the sum (number of people cross-subscribed)/(users in A) + (number of people cross-subscribed)/(users in B). So if a lot of people in a tiny subreddit are cross-subscribed, they get a big boost from the first term, but almost no boost if they make up a tiny sliver of subscribers to reddit B.