I've now written a fairly sophisticated scraper for Eliezer's blog posts based on lxml, which

• follows the Author links in "Article Navigation" to fetch all articles
• fetches and parses all articles
• identifies the title, body, and date
• fixes hrefs to internal references where possible, including where the reference is to Overcoming Bias and redirects back to Less Wrong.
• fixes all the weird Unicode characters as best I can where I can make a plausible guess
• finds and adds the forward references in all blog posts
• caches all network operations in a very simple dumb way
• writes them all out as very simple HTML with a very simple HTML contents page, in a form that Calibre works well on.

I'll share the script when I have time to sort out publishing via Mercurial, or email me if you'd like a snapshot copy - paul at ciphergoth dot org.