Well far be it for me to defend the JAA, but there are some fairly serious holes in that paper, most obviously the lack of any tests of statistical significance. The one vaguely interesting finding (that recency related accidents dropped by 40% over the intro of JAR) is glibly dismissed in the conclusions without any statistical analysis.
As the paper itself points out the purpose of JAR was harmonisation not safety.