Be careful about selection bias in your empirical data.
Suppose you're 50. You probably know more people 10 years away from your own age than 40. So no surprise you see more people die at 60 than at 90, because you know more at 60 in the first place.
And of course there's always a survivorship bias in life expectancy numbers. Life expectancy for a currently-living baby-boomer is over 80, because the boomers who die under 70 are already dead and no longer pulling down the average.
Suppose you're 50. You probably know more people 10 years away from your own age than 40. So no surprise you see more people die at 60 than at 90, because you know more at 60 in the first place.
And of course there's always a survivorship bias in life expectancy numbers. Life expectancy for a currently-living baby-boomer is over 80, because the boomers who die under 70 are already dead and no longer pulling down the average.