I think this went something like this (all numbers mine… it’s not about the math…)
Larian expected 3-4M sales, experience tells them 10% of that is during EA.
They end up with 2.5M EA sales, so now they wonder
1) is the estimate correct and the long EA phase resulted in many more people picking it up before release
2) does the industry average hold and there is a LOT more demand than they anticipated
3) a mix of the two
They probably concluded 3) and revised estimates to, say, 8-10M for an EA rate of 20-25%, but kept that close to their chest to not look foolish when it turns out to be scenario 1), only to find out it was actually scenario 2)….