I think tags might be getting parsed because we place the clips into the manager page, in order to pass it from the server to the client side code. So, IIRC, Redactor is not what's parsing the tags, and I don't think we really have a way to stop that.
Cases like this, where you have both textual content and other types of content element, is where you can really benefit from using ContentBlocks.
You'd have your text block, then a graphic block where the editor can actually upload their chart image and add the caption and anything else you need there, and then other blocks, together making up the page. All the while the generated markup is kept separate from the entered values.
Plus, if you do need snippets or other special tags, the user editing the page doesn't have to know what it is. They just select a field for their usage and add the content to it.
I know that doesn't necessarily answer your question directly, but I hope you might find it useful advice.