Ok, maybe this isn't a bug per se, but something that could be optimised; the input scheme for greek and hebrew is meant to be phonetic but it breaks down with the silent consonants.
So for example h:elohim doesn't come up with the correct lemma, it wants h:alohim...
I tried to do a search for a word beginning with an ayin eg עֶ֫בֶד
It would make sense to me to input that as eved or ebed but I couldn't work out what letter it wants for the ayin (I never did bother much with transliteration schemes, it seemed fairly pointless once you could read the script [:)])
I realise this is probably a result of the the transliteration schemes having characters for the silent consonants, but you guys are smart and I'm sure it wouldn't be too hard to work around this - if the first input letter from the user is a vowel that's fairly suggestive that the word has a silent consonant at the start [:)]