Well, to be fair, there are a couple of firm constraints on this tool.
First, the data-set it looked at was best-selling works of the last fifty years (more or less) and a control group of works written by the same authors. So it only predicts how attractive the title will be on a work by a bestselling author of the last half century.
Therefore, titles like "The Bible" and "The Quran" are evaluated as if some new book were published under those names. The Davinci Code is an understandable exception, the book made it entirely on hype and contraversy, it had nothing else going for it (other than the side-splitting but sadly unintentional hilarity, of course).
More importantly, it only takes the works of known best-selling authors into consideration. So if you aren't already a best selling author, then this tool is of little use to you.
Also, the titlescorer has only been tested (indeed, can only be tested) against the exact same data set used to generate it in the first place. You need to plug the title of a best selling book into it to check the validity. Even at that, it only claims 70% accuracy.
I think that it would be more fun to find out what the rules are by which it scores a title, but I don't see a link to that information. Perhaps google....