Like I said, it has to do with nothing but the difference between certainty and probability. It is the difference between knowing Schrödingers Cat is dead/alive, believing it is and having no position. Only the claims to knowledge represents firm statements; the beliefs absent certainty, like the absent belief, are only special cases of the general one: Uncertainty. Any degree of probability represents some uncertainty; only certainty represents certainty.

Claiming absolute certainty when none is impossible is very BAD for metaphysics, and physics remains ultimately just a special case of metaphysics. It is an empirical concrete metaphysics, and works quite well if we accept concrete empirical premises. Unfortunately, those premises are no more absolutely proven than any other (though a greater preponderance of evidence exists for them than nearly any other) and are practically useless in the real areas neither concrete nor empirical.

Unswerving belief (i.e. absolute certainty, especially absent absolute proof) is, as you say, bad for science. Atheism is such an unswerving belief; saying, "I believe there is no deity, and will live my life accordingly, but I might be wrong, and will reverse my position if (and only if) presented sufficient evidence of that," is agnosticism. That, and the impossibility of proving a negative, is why agnosticism is the rational position absent convincing evidence for a deity. There is a reason widespread erosion of religious belief did not instantly and automatically lead to widespread atheism: Because knowing my hand does not conceal a black pawn is not the same as knowing it DOES conceal a white one, or nothing; either is possible, as are many other possibilities beside those.

"Strong/weak" or "positive/negative" are far more sensible labels for agnosticism than for atheism. Agnosticism admits a gray scale atheism does not.