I was not trying to suggest that the example had any validity outside of the constraints listed in the example (which said the label run was faster with only one table). It was an attempt to demonstrate the difference between denormalization and a fully normalized design that is misread as being denormalized.