The atomicity of the internet - sealed documents at URLs - is really regrettable. A much more powerful design (although admittedly more complicated) would have been a "turtles all the way down", IPFS-style approach where any asset could be transcluded or linked into. But I don't think we'll ever see it.
By linked into, are you saying by URL? I notice that archive.is actually adds its own links to the side like a ruler. I think the status quo of linking only to headers is okay, but maybe you're envisioning something like linking to specific words in sentences. That would be great for quoting but the HTML would probably look/load horribly. Maybe there could be a system that let you link to offsets so it wouldn't need actual anchors in the source, but those could break with a simple addition of a word at the beginning. It's an interesting problem for sure.