Parallel to this fork, the original author has refactored his version, adding a minor feature and some performance improvement (avoiding duplication of effort between tab and space comprehension). Along the way, he made the actual calculated results worse, but rebasing the new algorithm onto the refactored code should not be difficult.
I should consider doing so once competing priorities have subsided.
Given that interactions with the original author have been rather unpleasant, and that users could end up getting confused between the two projects, I should rename this one to distance and differentiate it from its origin.
It probably wouldn't hurt to also include some report on methodology validating improved accuracy, or at least some (sanitized) examples of the files that work with this fork and not the original, so hapless users have something on which to base a decision. Of course they'll be a lot harder to find again now that they all work. :/