Resolved at last. It appears that the TBX had somehow got its template assignments messed up (and wasn't exporting OPML) and was using the 'plain' built-in template which is understandably why the Scrivener import failed in John's test. I had been worried there might be an issue the Scrivener end but finding this export configuration glitch, which was user - me not John - and not app error, all seems to fall into place given that the specimen OPML export did work.
The glitch didn't show in testing (for reasons perhaps for a thread another day) before upload but was immediately obvious on re-downloading it and looking at the export in HTML view and in the exported source code. Of course, for John who's not looking at source code the glitch simply wouldn't be apparent, so confusion isn't surprising on his part. The same TBX file, re-tested and re-uploaded again(!), now produces the expected OPML which already I know from John's report
does work. Test complete!
John, please throw away your existing OPML-to-Scrivener.tbx and
re-download it. You will find it will now work.
Wow, all this from one missing '/' ... and in misdiagnosing that a lot of false trails. But there at last

Anyway, this thread is now locked. Please start any new OPML discussion in a new thread (the "Exporting from Tinderbox" sub-forum seems the most logical place).
Meanwhile, I'm working on some OPML export example files for both inexperienced and more seasoned users. It may be a few days to get done; 90% of the time is the explanatory notes, etc. Until then, today's
revised TBX should suffice.