The really funny thing that you'll be wondering later is... why does jela get acc. jel and how do you differentiate it from the acc. of jelli? 
*mind blown*
We have actually discussed the word pair in question with David (
http://www.dothraki.com/2011/09/long-or-doubled-consonants/#comments). They truly become homonyms when inflected in accusative. That apparently happens sometimes. Happens in Finnish too, so my mind was luckily prepared.
Per the text given, these all make sense, but the last one. zhalia does not end with a lone vowel. Why does it get the /-e/?
Zhalia is a relatively new find. The way I see it, all other inanimate nouns we know have a stem ending in consonant. Some get a random vowel as a suffix for nominative, some don't, but the stem is the same type. Zhalia has the random vowel suffix too, but when you drop it off, the stem still ends in a vowel. This must be addressed in one way or another, and unsurprisingly it's addressed by an epenthetic /-e/. I say unsurprisingly because a) having accusatives that end in random vowels would be very confusing; b) losing the last syllabe without a trace would kind of mess the shape of the word.
Also: some consonat sounds (ɣ-phonemes at least) along Dothraki's history have become silent, which is why there are so often strings of vowels each pronounced in a separate syllabe (eg. leɣi became lei), so it's likely zhalia has been zhaliɣa.
I wrote the text given in the wiki. It's that "If the last syllabe of an inanimate noun is a lone vowel, and would thus be completely lost in accusative, an epenthetic /-e/ will mark it." that was meant to address the situation. I see now that I fumbled it. "Lone vowel" was meant to mean "a syllabe containing only a vowel" but is more likely to be be understood as "a vowel following a consonant", which is very much not what I meant to say.
