A classic statement of the issue is in Cleanth Brooks's "Criticism and Literary History: Marvell's Horation Ode," available on line at
http://www.hu.mtu.edu/~rlstrick/rsvtxt/brooks1.htm
His argument is summed up in this paragraph, I think:
But I propose to deal here with a more modest example than Milton’s epic. I propose to illustrate from Marvell’s “Horatian Ode.” If we follow the orthodox procedure, the obvious way to understand the “Ode” is to ascertain by historical evidence—by letters and documents of all kinds—what Marvell really thought of Cromwell, or, since Marvell apparently thought different things of Cromwell at different times, to ascertain the date of the “Ode,” and then neatly fit it into the particular stage of Marvell’s developing opinion of Cromwell. But this yields at best only an approximation of the poem; and there lurks in it some positive perils. For to ascertain what Marvell the man thought of Cromwell, and even to ascertain what Marvell as poet consciously intended to say in his poem, will not prove that the poem actually says this, or all this, or merely this. This last remark, in my opinion, does not imply too metaphysical a notion of the structure of a poem. There is surely a sense in which anyone must agree that a poem has a life of its own, and a sense in which it provides in itself the only criterion by which what it says can be judged. It is a commonplace that the poet sometimes writes better than he knows, and alas, on occasion, writes worse than he knows. The history of English literature will furnish plenty of examples of both cases.
A poem "provides in itself the only criterion by which what it says can be judged." It can be elucidated by external materials. If a poet refers to "the swan of Mantua," it seems fair to go outside the poem to find out he's alluding to Virgil. As Brooks says, "To put the matter into its simplest terms: the critic obviously must know what the words of the poem mean, something which immediately puts him in debt to the linguist; and since many of the words in this poem are proper norms, in debt to the historian as well." But these are tools to find out what the poem says, not who the poet was. When Brooks concludes, "Was this, then, the attitude of Andrew Marvell, born 1621, sometime student of Cambridge, returned traveller and prospective tutor, toward Oliver Cromwell in the summer of 1650? The honest answer must be: I do not know. I have tried to read the poem, the 'Horatian Ode,' not Andrew Marvell’s mind. That seems sensible to me in view of the fact that we have the poem, whereas the attitude held by Marvell at any particular time must be a matter of inference, even though I grant that the poem may be put in as part of the evidence from which we draw inferences[,]" I can only agree, even though I must sometimes remind myself that knowing what I know about Pound's crackpot fascism and Eliot's semi-genteel anti-Semitism makes no specific poem either better or worse.
RHE-