It would require non-trivial work to scrape the pages to get the html, set up parsing to handle different types of content, get into a format that could be imported into something else, etc. Not something I'm willing to spend my time setting up, and what about all the people who might want to export their tumblr content but can't program.

For e.g. Wordpress to import Tumblr blogs they need to write an importer just for Tumblr's format, whereas if Tumblr used a standard like RSS, WP could use a generalist importer.

Don't underestimate the value of standard formats; the transport mechanism (XML/JSON/HTML) is a small detail.