I think his point here is that the effect in replication is closer to 0 than to the original claim. It might be more obvious if he chose an order of magnitude difference as an example - going from the dominant factor to technically-not-nothing might be replication but it's not vindication.