Except those estimates are assumptions, not based in the data. When they’ve gone back and canvassed populations for antibodies that would indicate significant populations of mild/asymptomatic cases, they are not finding the numbers you’d assume they’d find, the ones that would justify assuming a lower death rate.