At first sight it seems to me that the coverage being positivelly correlated with how unusual a death is and the number of people dying in a single event, would explain that graph.

I bet if we dig into the details of the Accidents class we would see a pattern were uncommon kinds of accidents and/or those with a large number of deaths ("man killed by falling crane", "plane crash") get lots of coverage whilst common kinds of accidents with few victims per event ("a car crash involving a single car") get a lot less coverage.