010101 see my post above. In @djalmasi case the # is added by Secure HTTPS itself as part of its HTML, so it will basically happen anytime Secure HTTPS replaces an image. In my simpler example clarkwinkelmann the # is added by myself as part of an alt image attribute.
I doubt this can be fixed just by editing the regular expression. It's impossible to "parse" HTML via a regular expression to isolate text nodes vs markup.
My suggested solution without re-writing how the whole extension works would be to "walk" down the HTML nodes and run the regex on each text node that's inside a paragraph, instead of altering the innerHTML of the whole paragraph.
