Do you really think that plot is complete bs though? Sure it cant be 100% accurate, but it seems to me if the "stamp collecting" effect were so strong you wouldn't see such clear trends. Also, wouldn't you think that such macroscopic analysis would account for the records incompleteness without necessarily destroying the truth of the trend? I tried to see how they derived that plot from the original article but I cant say for certain yet I know. Here's the link though https://muller.lbl.gov/papers/Rohde-Muller-Nature.pdf