Yes, I completely agree with you! You will have much less variability.
However, that's not what the paper argues. The paper argues that you would have more variability.
I'd appreciate you taking a look at the original paper (specifically, the bottom of page 2 [edit: and Figure 1]) and understanding my original argument.