It's definitely relevant here on HN -- a good article with a bad title probably won't go anywhere. A good article with a great title will stay on the front page for hours.
Also perhaps a downside of not having Slashdot-style summaries, in that it encourages blogspam. If I find an interesting academic paper that I think HN might like, just submitting it by itself rarely goes anywhere, because there isn't space to explain why I think it's of interest to HN. Oddly enough it actually stands a better chance on Reddit, because you have about one sentence worth of space in the submission title to explain it. On HN, it's better if you reblog the paper with a one- or two-paragraph blurb, catchy title, and then submit that instead of the original paper. People will then complain that you submitted blogspam instead of the original, but most of the time they'll upvote it more than they would've upvoted the original despite the complaints.
The guidelines state:
You can make up a new title if you want, but if you put gratuitous editorial spin on it, the editors may rewrite it.
I frequently write better headlines for items I submit to HN and they are not often changed by the editors.