I don't know of anyone who would replace a drive because of a single bad sector. It's really strange that the extended test did not stop when it tried to read that sector. Maybe it's just marginal, and was read successfully by the test.
The way you fix a bad sector is by writing to it, which will cause it to be reallocated if it's actually bad, or else removed from the "pending" list if it can be written successfully. The
Bad Block HOWTO has instructions for doing that, but that's all predicated on identifying the bad sector. Unless you can get the test to fail or see an I/O error from a "read" operation, there's no way to know where it is.
I suppose you could run "
dd if=/dev/sd{X} of=/dev/sd{X} bs=64k" to copy the whose disk back to itself. I really don't like that idea because of the chance of a memory problem or other glitch sliently corrupting the data, but I can't think of any other way to rewrite a bad sector that resists being identified.