So I think you are drawing a distinction between doing what you expect to lead to the greatest increase in utility, and doing what leads to the largest mathematical expectation increase in utility. You're right that the latter is not actually possible, and the former could possibly lead to worse outcomes than following the rules. But this is not a stable equilibrium.

1. If you discover that you're so bad at predicting the consequences of your actions that always following the rules would produce better outcomes, then the good utilitarian thing to do is to give up on your own judgement and follow the rules.

2. If you discover that people you convince to follow utilitarianism end up in situation #1, then the good utilitarian thing to do is to lie to people about your beliefs and convince them to just follow the rules.