Thanks everyone!
The problem has been solved, and the odd characters are gone.
The query didn't work without some modifications. I ended up using
select xwd_fullname from xwikidoc where xwd_content like '%' || chr(1) ||
'%';
(concat() only takes two operands, so it was easier to use ||. My sqlplus
tool didn't like "%" and char(1), so I had to use '%' and chr(1).)
After this character was removed, I got complaints about 0xf. The page was
identified with the equivalent query, but using chr(15).
I also had to remove some of the old versions from the page history. Oddly
enough, the queries still report that the odd characters are present in the
database, but exporting the entire wiki works fine now, so the odd
characters are no longer a problem.
By the way, the text with the odd characters had entered the wiki when text
had been copied an pasted from UNIX terminal windows.
Agan, thanks everyone for your invaluable help!
Peter