Pay attention to what I said BB2K..
I gave you the hottest T° in copper so point in the middle of the die BUT T° distribution isn't equal on the die , it's "concentric" so a T° average (less than higest T°) must be calculated on whole die surface (as BillA die it's not T° constant on the die, in reality it's different because of components into the die, L2 heat more than other die units so T° distribution is "random").
This average could change between 2 blocks... The 4° are not real because it's an savage extrapolation (?) to 70W to 100W I done in my mind
(not the real geometry too)
Take care about T° because all is "perfect" in a simulation (no TIM, perfect contact, etc..), depends on my reference too...It's a prediction !!