The problem with that being that the subnodes wouldn't necessarily be distuinguishable using text to identify them.
For example here's a simple XML file:
<title>The Big Book of mIRC</title>
<content>Wow, what a great book</content>
<title>Monopolising The Marketplace</title>
<author>B. H. Gates</author>
<content>Blah blah blah...</content>
<title>XML For Dummies</title>
<author>W. W. Web</author>
Now, say you wanted to get Arnie's review of 'The Big Book of mIRC'.
Immediately you've got a problem in that all the children of 'library' are called 'book' - no simple text identifier available. With a language that had arrays etc. that wouldn't be a problem, it would simply be a case of referring to it as library/book (assuming a 0-based array). With mIRC there's no such luck, using traditional mIRC syntax the way to identify that would be something like $xmlsub(library,book,1) - which doesn't look too bad.
So then we move on to getting the correct 'review' subnode, again there's more than one node of the same type, with the array-enabled language it's just a matter of library/book/review, but what would that look like for mIRC? The previous $xmlsub usage is useless since it wouldn't return an object, just $true. How would be refer to it? I suppose it could simply use some syntax which resembled the array style, but then it wouldn't follow traditional mIRC syntaxing and could be misleading to a newcomer who might think he was dealing with actual arrays.
I'm not trying to put down the idea of XML functionality in mIRC, I use XML myself a fair bit in webdev, it just seems a bit of a stretch at the moment.