Is a selfish advertising attack with bad radio waves (γ = 0) still profitable?

The 2013 selfish mining papers of ITTAY EYAL and EMIN Gün Sirer (0) introduce variable γ.

We indicate the ratio of different (1-γ) of non-full mines at a point different from the ratio of the honest miner to select the mine in the swimming pool block.

The basic idea is to withhold blocks until the attacker grass is learned about the competitive block of honest miners. Honest nodes prefer the first block they see, so this variable shows the ability of an attacker to ride the block in front of an honest block.

Specifically, this represents a part of Miner See the attack block first. As described in the simulation section:

We assume that block spread time can be ignored compared to mining time. In the case of the same length, we artificially divided the miners without grass, and the ratio of the γ of the pool and the remaining mine are at a different point.

Simulation creates Figure 2.

I am confused about the most pessimistic scenario γ = 0. Intuitively I will expect such a swimming pool to lose money consistently. Why doesn’t the red line stay under the gray line?

Is there an additional assumption in the paper I missed?

(0) https://arxiv.org/abs/1311.0243