It seems hard to take that interpretation at face value (20% seems to be an effect of a week 1 crack post-release with total revenue lost estimated at 25%; week 3 crack has estimated total losses at ~12%, and week 7 crack at less than 5% of total revenue loss..., ~0% for week 12+ cracks).

This is also based on extrapolation on top of extrapolation covering only 86 games with "majority" surviving without cracks into week 12 — how significant is the effect if there are only a few games with cracks in early weeks (if it's 43 games across the first 12 weeks, it's less than 4 games per week on average)? How big are their revenues and copies sold in absolute numbers? (I do not have access to the full paper, perhaps it's answered there)

But to be precise, even if all of the above is covered, this is not proof that DRM increases sales, but that crack availability for Denuvo-protected games decreases sales depending on the timing — it is a subtle distinction, but perhaps publicity of a crack availability motivates more people to take that route?

Finally, let's not forget that game companies care about the profit (and revenue is only a proxy): looking at lost sales does not show how much a studio can save by not investing in DRM protection and thus having a higher gross margin or cheaper price to entice more customers.