Does one or two very, very old people make much difference for aggregate statistics that lumps people 65+ together already?
I made the death rate slightly less for the calculation to maybe fudge out the cutoff age a bit. I.e., I moved the probability one death-age from 104 to 105, and slowed the curve a bit. Here are the two actual fits (exponential, and logistic):

I could redo the calculations with a higher cutoff and the logistic. (I want a cutoff to maintain finite vectors. But it could be 150, or more.)