They may have accepted some in the past, but I've never seen it, and I've seen a lot of PRs closed unmerged.
-1 closed by the author realizing it had already been done
-2 closed saying they were going to fix it differently
-1 closed saying they took some code from the commit and fixed part of it differently
-6 merged
seems quite reasonable to me (albeit this is only from the newest 10 closed pr obviously)
Glad to hear this seems to be changing; 6/10 is pretty good.
They may have accepted some in the past, but I've never seen it, and I've seen a lot of PRs closed unmerged.