Well, the meta-level of what you said is "Updating beliefs when evidence is against them is not always beneficial." I think there are articles here that challenge this kind of meta although I cannot point to them, I am fairly new here. But I still see the issue namely how exactly do you decide, by what algorithm, what other beliefs of yours you want to update when evidence is against them and what not? So it seems you will have to competing motives, to execute the truth-seeking algorithm and the belief-defending one and they may weaken each other. Yet, I think with some compartmentalization it can work but it may be difficult.

To put it different words, you can simply put a taboo on full-on truth-seeking wrt religion and let the truth-seeking algorithm run elsewhere, but you have a reason, an algorithm for that taboo, maybe not fully conscious and that may conflict your truth-seeking algorithm in other fields in more subtle ways: perhaps not handing out an clear obvious taboo, but biasing results. Or to be blunt: non-rationality has reasons and methods too, and thus leaks out from compartments and contaminates.

Just my 2 cents, I am also a beginning learner here.