Run this in wxMaxima:
plot2d([parametric,2-2/(%e^-(%e^-t)+1),((2*%e^(-%e^(-t)-t))/(%e^(-%e^(-t))+1)^2)/(2-2/(%e^-(%e^-t)+1)),[t,-2,5],[nticks,80]])$
Basically it fixes Hubbert Linearization by using a skewed, rather than normal distribution on the production curve. It fits the early data points a lot better.
Moreover, the lower 48 production curve is, indeed, skewed in the required way:
