First, there are limitations. That’s why I brought in the comparison with InDesign. Indesign lets you control everything, yet it’s already very hard to control the flow of elements. So it’s only natural to expect less from such a system that is HTML+CSS.
You have to set expectations. A complete perfectly working automatic grid system is a Graal. Read: it has never been made. But for every goal is an answer. And by working it the right way, very nice results are always achievable.
What I have a ver very hard time to accept is “RT element in Webflow is clunky”. At best what I’d be OK with it “Rich Text is a difficult sport”, and “Typography, especially web typography, is a very complex thing.” I’d ditch “Prepare to be annoyed” for “Prepare to be challenged” 
I’m doing web stuff professionally since 1997. I’ve seen RT element come in the scope. I’ve struggled with them. I yelled at the ones we used on Drupal, the one we used on Joomla, I’ve seen the light with some of the ones used with Wordpress, only to be disappointed later on. Webflow’s RT is best in class so far. Limited, yes, not full featured yet, yes, but quite complex behind its apparent simplicity. Having put hours on just styling the figure group of elements, I have yet not reached the end of it. And all of this without using any pseudo element because they’re not there yet.
I just read your last answer: seems you never had a chance to style a <figure>
element. You can do it by selecting an image, go up in hierarchy to the figure, force the styling of the figure element, limit the effect of the styling to the rt’s class.
Here you can see a figure element losing it’s wrapping when switching to portrait mode
Rather than adding classes inside a RT, the workflow here is to add class to the rt, then force the style of base elements within, and limit to the rt class. So all of this is usable in a CMS environment. That’s true for titles, paragraphs, lists, figures etc