Okay, I don't think this question is substantially different from another question which I've already marked, but I'll do my best to give you an answer here anyway. Simply put, you're misunderstanding the difference between Act utilitarianism and rule utilitarianism.
I will define act utilitarianism and rule utilitarianism as follows:
Act utilitarianism:the position that in order to maximize some desirable feature (or minimize some undesirable feature), each agent should for each action pick the course that produces the optimal outcome on the basis of his or her individual knowledge and expectations.
Rule utilitarianism: the position that in order to maximize some desirable feature (or minimize some undesirable feature), rules should be established that produce the optimal outcome across many actions incorporating the limitations and constraints on the information
available to those agents.
Thus, both seek to optimize outcomes, but act utilitarianism does so by having agents optimize their own acts and rule utilitarianism does so by optimizing rules for actions (not generally at the agent level).
You propose the following rule:
Rule Set B': Follow Rule Set B in every case except for in situation X , in which case do act A
What's proposed in B' is basically a reduction to the following rule: always be an act utilitarian because this produces maximal happiness. I take it that's your point, viz., that the best type of rule a rule utilitarian could come up with is to be an act utilitarian. And that's probably true on the definition you supply.
But the definition you supply differs from the one I suggest here precisely in that it does not incorporate the problems with calculating utility on the spot that rule utilitarianism seeks to address.
Rule utilitarianism exists primarily because act utilitarianism on the view of many suffers from an epistemic problem. This is inherent in the formulation of Mill (and Bentham) which does not spell out an answer to the following important question:
*Am I optimizing actual or expected consequences? (and do I know with certainty the outcome and effects of my actions? do I have enough knowledge or depth of processing power to calculate correctly?)
In other words, am I doing right if I do what I think produces the most happiness or am I doing right if I do what produces the most happiness?
Both horns are problematic. If I pick the think horn, then Mao killing the sparrows occurs. And Mao is a great guy there for wanting to supply enough food.
Conversely, if I pick the produces horn, then it's not clear that my choices matter for ethics. My accidental choice to not buy the cake on discount may lead to great happiness for a poor family even though I never thought about it. Or my dropping change on the floor might lead to a horrific accident. Here, we'd have a very strong form of moral luck.
The rule utilitarian hedges the question a bit better by coming up with rules that are designed to optimize said goal but that don't depend on the individual agent arriving at the epistemic best path under the stress of the moment. It doesn't completely eliminate the horns, but it dilutes their painfulness, because it can suggest that our duty is to follow rules that seem optimized to produce the ends we want even if in particular cases they don't.
Thus, we can return to your reduction. The rule utilitarian rejects it because it doesn't sufficiently grapple with the agent's epistemic realities. A major point of transitioning from act to rule is to get away from the vagaries and gambling that arise from making judgments at each particular juncture for each particular act.
Thus, for the above reasons, I don't think your reduction truly captures the position of the rule utilitarian. Instead, it just makes him an act utilitarian who doesn't think rules add anything.