"Life insurers can predict when you'll die with about 98% accuracy."
This conclusion isn't supported by the linked document. The document instead is talking about expected vs actual deaths among demographic groups as a whole, not individual people. And that expected vs actual is just history + trends. This doesn't mean that insurance can say that Joe Blow is going to die in June of 2027 with "98% accuracy", obviously.
Will you be one of them? Click here to find out!
Men under 40 essentially have incredibly low mortality. Once you exclude cars and suicide it drops basically to zero.
Even things that are very very bad for you, like being very obese (BMI>40), are factors that scale a mortality that is massively dominated by age and sex. Even 3x mortality at 25 rounds to zero. At 60 background become 1% per year and a 3x increase would make a huge difference.
I would wager $50 that there are no contiguous collection of zip codes in the USA where the average 40 year old has a higher life expectancy than the average 30 year old in any other contiguous set of zip codes assuming there are over 1000 people of that age in both.
Pretty easy to predict if you're willing to make it happen.