There's nothing that surprising here, actually.
The RM pages have a decently high signal to noise ratio (like the rest of XF, really), and the notion that Google can, shock horror, examine the page contents to find something specifically identified in the source as a rating value... should not be a surprise to anyone here.