There are deficiencies with the current matching system. It doesn't work in relatively dense fields. It seems to be slow. A more brute force method than the current algorithm seems to be both faster and more robust (ask Colin S. for details).

The result should be a faster and more robust implementation of a one to one matcher.