He briefly addressed this in one of the appendices:

"Another reason to sac bunt (or bunt in general) is that the tendency to sometimes do this induces changes in defense which make non-bunt plays work better."

It apparently changes the probabilities slightly, but not enough to change the overall conclusion.