For further clarification, Liddell & Scott's Greek-English Lexicon (9th edition) lists the correct spelling as a lemma and adds "freq. written corruptly [Brewer's spelling]". L&SGEL also includes the corrupt spelling as a separate lemma, with a cross-reference to the correct variant (although it does not cite instances). Glare's Oxford-Latin Dictionary also gives the two versions. [I don't have Lewis & Short to hand.] So it appears that this problem originated in ancient times, which has persisted. The OED includes an etymological note about this under the adjectival form of the correct variant; everywhere else it uses that variant, except in three 17th century citations, where the corrupt version was used in the original. The modern Greek name for the theme word also has two variants, according to the Times Comprehensive Atlas of the World.
Incidentally, it is amusing that there is a connection between this and one of the clued lights, as hinted at by Stavrolex ...
I would agree with those who interpret the preamble as suggesting that the corrupt spelling should be submitted, were it not for the fact that the BRB lists the correct version under another headword (i.e. confirms it, as the preamble says). I would expect both versions to be accepted. In other words, I agree with Stavrolex on all counts (except perhaps for assuming that Zag and the checkers realised what was going on). After all, we are etymological cousins :>)