First point: there's a useful distinction to be made between 'translation' and 'imitation'. In the seventeenth and eighteenth centuries, imitations were all the rage. The source text crosses over into the target language and undergoes a number of creative metamorphoses. Its images and allusions are often updated. New images and ideas and sometimes whole new lines may be added ad libitum. Robert Lowell did much the same thing with his versions of famous poems in other languages. This is always a perfectly legitimate exercise, even if more literal and accurate translations have been consulted along the way: legitimate, at any rate, to the extent that what results genuinely adds to the target language's literature something that is of lasting value.
Second point: unfortunately, many of the best poems in English purporting to be translations from other languages have been made by poets whose grasp of the source language was either shaky or virtually non-existent. Pound's magnificent slim but seminal volume Cathay commits not a few risible howlers, but has outlasted many much more accurate translators by tin-eared scholars surrounded by the very best reference books and chock-full of academic learning, but totally devoid of any feeling for their own language. Rexroth began his career as a translator by working, not from the original Chinese texts, but from Judith Gautier's translations into nineteenth-century French. (Admittedly, his Chinese improved over time, as did Pound's.) I doubt whether Edwin Morgan was fluent in all the languages he translated so skilfully and powerfully; but who would be without any of his work as a 'translator'? Fitzgerald's Omar Khayam came to us unfiltered by any depth of Oriental learning on his part. Many of the best poetic 'translations' in the English language are by poet-dilettantes of this kind.
This is not to deny that some of the best translators are also those whose scholarship equals their creativity: Arthur Waley and Burton Watson spring to mind in the field of translation from classical Chinese. But more frequently, the translator whose knowledge of the source language is most advanced lacks a feeling for how the target language works.
It must be remembered that whether your command of the source language is superficial or profound, any translation is always an interpretation: in much the same way that an instrumentalist doesn't just perform a composer's score, but also interprets it: and every interpretation is (as it should be) different and unique. Some will be better, some worse---but all are arguably necessary: since one can learn as much from a poor interpretation as one can from a good one. One can take this analogy one step further. Some classical musicians choose NOT to listen to other interpretations while preparing for a performance or recording. Others listen to as many as possible. I don't see why a translator should impose upon him- or herself a self-denying ordinance of this kind; personally, if there are other translations around, I'll often seek them out, if only to compare them with my own. Sometimes they'll influence this or that draft, and sometimes they won't. If they do, I'll either consciously alter a word or phrase to ensure that it DOESN'T resemble the other translation, or else sometimes I'll think, no, that particular word-choice is so much better than mine that I'll borrow it, because I don't see how it could have been done better. It's certainly more honest to acknowledge this in a footnote, but if only one word is involved, I don't think it's in any way obligatory. As one contributor to this discussion has already pointed out, some 'plagiarisms' occur through sheer absent-mindedness: one consults so many sources while making a translation of one's own (which can be a very protracted process, sometimes stretching over years and even decades) that it's easy to forget which bits were entirely one's own and which bits weren't. (But it's better to check before going into print, I agree.)
Third point: it's true that some 'translators' feel compelled to start their own versions of a poem they've initially encountered only in someone else's translation precisely because they're so dissatisfied with it as it currently stands. Yes, what follows is something that's been 'edited', not translated, and (unless you're already a very great poet in your own right) it definitely shouldn't ever be printed, or submitted for a competition. But it can still be a useful exercise, if only in that it might lead one towards a language that one hadn't previously bothered to learn or get acquainted with.
Finally, literatures often actually benefit from misunderstandings and mis-translations. One example. Ezra Pound and Amy Lowell founded the Imagist movement on what is really a gross distortion of what Chinese poetry is really like. (Short, metaphor-free 'snapshots' of reality? More frequently nothing of the sort. Many Chinese poems are extremely long, and are stuffed with 'show-off' allusions and wildly opaque metaphors.) But the influence of Imagism is pervasive in British, American and European poetry a hundred years later.