This seems like a procedural step back from XSLT.
Don't get me wrong, I certainly am not fond of XSLT, its programmatic fundamentals were certainly lacking (technically there, practically a pain in the ass) and the implicit nature of the "current node" always weirded me out, but the very descriptive match/transform of XSLT was reasonably great, any not-too-badly-written stylesheet would show the output structure shine through.
If I wanted something to replace XSLT, I'd take adding XLST-type (match-and-transform) capabilities to a general-purpose language, not attempting to extract a fucked-up version of a jQuery DOM manipulation chain.