The statistic looks appallingly bad to me viewed as a poisson process. If you had thousands of planes with zero crashes over decades would you consider that as zero data points?