My guess would be because certain combinations of factors should change the way other factors score (i.e. Non-Religious Gay may be of neutral value if one is white, attractive, educated, and from an affluent family, whereas if one is from a poor latino family, it would be something like -250).

But that kind of thing needs a calculating database, not a simple chart. There are a lot of problems with this chart, but it gives an interesting jumping-off point for discussion.