Translation would seem to be an "unoriginal" form of writing. What Jakobson called "intralinguistic translation" is paraphrase, or the writing of a text that translates within the same language. The only thing more unoriginal than that is transcription. We know from Pierre Menard that transcription can itself be "original."
(Paraphrase does differ from translation, though, in that, curiously enough, it cannot be as exact. As Carlos Piera points out in "Sobre traducción, paráfrasis y verdad," we can translate César Vallejo's word "oxígeno" with a high degre of exactitude as "oxygen." But to paraphrase, we need to find our own words.)
I'm assuming that translation is original because, for example, we cannot simply take a translation, change a few words, and call it our own. Even for a literal translation, we assume that the translator can claim the translation as her own, copyright it, complain if someone copies it verbatim (or even almost verbatim.) Translation might look like semi-plagiarisms of one another by coincidence, of course. There may be one solution to a short line or sentence that occurs independently to various translators.
So called "creative translations" will, of course, look different from other translations, by the same measure that they differ from the original. So one way of looking at the problem would be to say that the originality of the translator lies in not always using the word "oxygen" when that is the most logical choice. In other words, translation as paraphrase and interpretation. In a way, a paraphrase depends on an understanding, whereas a translation can be a transfer of ill-understood signifiers (if we follow Piera's logic.)
The translation of poetry results in more poetry. In other words, the translated text stands before us as a poem. The poet is the translator. So instead of saying that we're reading Adonis in translation, we should way that we are reading poems by name of translator, poems that just happen to be translations of other texts that we are unable to read.
That this is not satisfactory solution suggests that our notions of authorship come into play. What we attribute to the original poet, in a translation, is the poetic self itself. Poetry is (usually understood as) the expression of a self, of personal experiences, even if the speaker of the poem is not (easily identified with) the biographical poet.
So the standard model of translation is that the two functions of the poet are split between two authors. One, the translator, provides the poetry itself, where "poetry" is understood as the formal structure, the use of language, etc... The other, the "author," provides the self, the experience. The translator would be the ego-less writer, the one without self or experience, the writer with nothing to say. Not surprisingly, this resembles the function of the ghost writer, or what in Spanish they call, a bit racistly to my ears, the negro.
The problem is that this split seems to reinforce an "unpoetic" way of thinking, since it separates "form" from "content." Anyone who claims to think these are inseparable should have no business closer than 100 feet from the translation of poetry.
One way around this is the translation of poetry where nothing much is going on. In other words, the translated text itself reads as flat prose, without any charge or electricity. The translator might say that the translation was easy, because the original is like that too. I have noticed this a lot in translations of Eastern European poetry. This poetry is read, indeed, for its content; its political positions or value as "human" testimony. I don't get much out of this kind of poetry or poetry-translation. I'd rather read a memoir if what I'm after is the recounting of personal experience in narrative form.
Another way is for the translator to be a poet and use his or her own "poetry" as the poetry of the translation. When we want to read a translation by a famous poet, we want to do so because we think that the target language text will be good in the same way that famous poet's poems are good.
What is more, what we really want to see is a kind of meeting of two great poetic minds. In other words, we want to see poetry 1 (the original poet's genius) meeting up with poetry 2 (the translator's genius) in a marvelous way. This will be an original event because it has never happened before between these particular two poets. Celan translating Dickinson, say. Now this might not always occur, or we might over-value the result. While I appreciate that Ashbery has translated Rimbaud, and have nothing to complain about in his translations, I don't think I feel any great frisson there. Without the name of Ashbery I would simply see these as competent translations.