I agree. Which is why I would argue that the material change of wording, even in an unchecked box, should destroy the presumption that the document is prima facie valid. While the presumption of validity may have an unfortunate effects in cases where the model is followed precisely, I would argue that in this particular case the fact that the model wasn't followed should void the presumption, rendering the broader problem irrelevant to this particular case.
It's a good argument. But it isn't a slam dunk, and the judge's decision won't necessarily turn on that argument. He has other tools to justify whatever outcome he wants.
Check out the Browning case ...
http://www.freerepublic.com/focus/f-news/1380586/posts?page=217#217