Saw a good idea in a presentation the other day, perhaps with limited applications though.
ylim <- 1.1* range(mod$residuals)
To expand the axis a bit and ensure the residuals have a bit more space either side that usual.
NOTE: This works for residuals given that they will always be both positive and negative, otherwise you might need:
ylim <- c(0.9,1.1)*range(mod$residuals)
which may seem cumbersome but could add consistency if automating the production of multiple graphs.