Originally Posted by wallcraft
When you click on "convert individually" there is a "look and feel" icon on the left. It has two options that might help. One is "input character encoding" and the other is "Transliterate unicode characters to ASCII". I suggested trying cp1252 in the encoding box, and what worked for monkeygirl351 was checking the transliterate box.
It can be hard to convert TXT to something else because plain text has no metadata (describing the properties of the document). The encoding is the convention used to map from the file's binary bits to characters. For English characters this does not matter because most encodings use the same convention for them, but they differ in how they map other binary patterns to characters. What you are seeing is binary patterns that have been mapped to the wrong character.
I know this is very old but I wanted to add that I used this technique to fix a couple fussy ebooks to mobi. They would read fine using view on calibre but in the kindle I got little question marks in boxes for many of the dash and quotation characters.