This question has more than just abstract philosophical interest: We presently teach science in government-funded public schools, where religion is explicitly excluded by the Constitution. If science is indeed a religion, then we ought not to teach it—indeed it would be illegal to do so.
The controversy comes to a sharp point in the teaching of evolution in biology class, but the issue has much wider implications. If evolution is excluded, why not exclude cosmology? Heliocentrism? Why not exclude all science as a preposterous attempt to establish truth using only human tools, without divine revelation?
If we are to apply rulings on religion to the teaching of science, it is clearly legally insufficient to merely allow students to opt out of specific science classes on an individual basis. This remedy has been decisively rejected by many appellate and Supreme court decisions because of the pressure the mere state endorsement of a religion has on the individual student. Congress (and, since the Fourteenth Amendment) state and local governments shall make no law establishing a religion. No law means means no law at all, not even a law allowing individual exemptions.
Science arbitrarily privileges a particular metaphysical system, a definition of "truth":
(ST) A scientific truth is the simplest logical explanation that accounts for a body of uncontroversial facts.You cannot prove this definition of "truth", and this definition is viciously self-referential: (SP) is not itself a scientific truth.
The question then becomes: Should we define "religion" to mean an arbitrarily privileged metaphysical system, and thus include science? Can we do otherwise?
Let's tackle the second question first: If we can't find a principled distinction between science and religion, then there's no point in talking about whether we should do otherwise. Religion and science are indeed similar on the metaphysical point; on the other hand, a similarity does not entail an identity.
If we examine the ideologies uncontroversially labeled as "religious": Judaism, Christianity, Islam, Buddhism, Hinduism and the like, an obvious pattern emerges: These ideologies say (in a predictive or prescriptive sense) nothing at all about the world of perceptual experience. More importantly, they either say nothing at all about objective reality, or they say things about objective reality (e.g. God really exists) that cannot be publicly determined by appeal to perception.
Physical science, on the other hand, is defined in terms of a foundation of shared perceptual experiences, and publicly available methods—logical accountability and simplicity—to determine truth from that foundation.
So on this basis, there is not only a difference between science and religion, they are mutually exclusive. We can make a principled distinction and label some arbitrarily chosen metaphysical systems as "religious" and some (including science) as "nonreligious".
It meaningful then to ask if we should make this distinction, i.e. is it in all our best interests to do so.
If we do not make any distinction between religious and nonreligious metaphysics, then it is hard to see how we can have any laws at all; indeed it's hard to see how we can assign any sort of public meaning to words at all. One might simply say, "Your definition of 'murder'" is essentially religious and it contradicts my religion; your laws against murder are therefore an unconstitutional infringement on my free exercise of religion. It's difficult to see a person of any religious persuasion willing to take this sort of nihilism to its logical conclusions.
If we are to have any sort of state which includes the religious clauses of the First Amendment, we can and must draw some sort of distinction between religion and non-religion. Simply erasing the distinction by philosophical sleight-of-hand will not satisfy anyone.
Tomorrow I'll talk about whether the distinction drawn here between science and non-science should be the preferred distinction between non-religion and religion, respectively.
First Freedom First
 Technically, you can prove (ST), but only at the cost of privileging some other unproven metaphysical statements.
 Phenomenological science is defined in terms of individual personal experience.
 One could always deny shared perceptual experience as a foundation; in practice, however, the only people who actually do so tend not to care about laws and society at all.