This is not right. We show that, even with no positional advantage in the network (i.e. gamma=0), the attacker can make more than its fair share.
For example, let's assume a "levels" topology, where mined blocks propagate in discrete turns from the miner to its peers, then to their peers, etc. If there are more honest peers on the first level, wouldn't the "honest" block (X in the paper) propagate faster than the pool block, P?
BTW, I found the paper absolutely fascinating regardless of this relatively minor issue. As you said, even with gamma=0, an attack is feasible.