I think there's a bug, if you put in a birth date in the 1930s for the USA, it says you have negative time remaining.
That a person alive now is expected to die in the past?
> Consider that at 91-92 years you will have exceeded the expected number of years for your cohort.
The correct method for estimating remaining life is using life expectancy at current age not life expectancy at birth, since you want to know how long one is expected to live given that they have not, in fact, died already.
At 92, life expectancy is 3.54 years for a man, 4.2 for a woman, not some negative value.
This seems an uncharitable interpretation. Instead, I expect that the value shown is the "expected age of death" - "current age", which if your current age is beyond the average results in a negative number.
FWIW and as you note, this isn't the most useful result since alive people can't die in the past unless we accept some sci-fi premises.
> The correct method
While we're on the topic... More correct? Many factors that would improve the prediction weren't collected and used but a better prediction would take those into account.
So... Yes, your suggested method seems more useful and accurate yet I wasn't going for that goal so much as noting that it made sense to be that a lazy/naive/PoC/first implementation might get a negative result using exactly the formula above.