CFM is only part of the story. You can actually have a cylinder head that is down on CFM then another head but kill it on the engine dyno.
Well said. From Tom Wilson's superb article on the Boss from a few years ago...
From here."So the exhaust port geometry turned out quite differently than what you would come up with on a flow bench. And so if you flow these heads on a bench, there are actually some lift regimes where these will look worse than a Coyote, whereas on a running engine they actually out-perform it."
The counterintuitive area was the exhaust port floor, right in the throat of the port outside the valve seat. This area was choked, and the computer "actually told us to open between the guide and the short turn, which we could only do by pushing the floor lower. This would have looked worse on a flow bench," but it definitely made more power.
Along the same lines, cutting material from the backside of the exhaust valve head helped because the team could see the airflow was more sideways, across the valve head, rather then the textbook example of bending around the head in a classic tulip shape.