I like your numbers better than mine (that were pulled out of my ass to some extent), but I do NOT know what is ristos calculations... so I guess, I am shooting from a position of disadvantage.
It's just an exponential curve.
Others have observed that the better empirical fit, on an R^2 basis, is to an exp(exp(t))+k model, which would put the numbers higher still. Not sure whether that "double exponential" fit is still better. It has to be deteriorating now.