Note: I don't like TXT as a source format, because it loses too much information (e.g. italics).
What works for me: Take a source format, whatever that is, and convert it to a bare-bone (X)HTML: chapters, paragraphs, italics (with classes denoting the type of italics, e.g. thought, foreign word, real emphasis...), blockquotes (again, with classes for letters, signs, poems...). That's about it as far as HTML tags go. Add a standardized header and footer. When done, convert all books at once using my H2LRF "preprocessor" into any target format I like.