This class of models is a simple extrapolation of historical aggregate lifespan time-series, and thus cannot model non-trivial pharmacological and genetic manipulations even in principle, unless these are a part of a large smooth sub-trend that has already begun.
One wouldn't be able to predict that some random flavonoid molecule extends lifespan of lab mice by 10-15%[1], and if we took the [overstated] paper's title claim uncritically, this research wouldn't happen. And yet it happened, and more than a dozen clinical trials in humans are underway.