It is a little confusing because we are talking about both rendering and GUIs, but ... "retained mode" in this case refers to the GUI itself, not the method of drawing. Motif and Xlib are "retained mode" in the GUI sense because if you want there to be a button, you instantiate that button and register it with the library, and then if you want it to become visible or invisible or change color you call procedures that poke values on that instantiated button. In IMGUI you don't preinstantiate; you just say "draw a button now" and if you don't want it to be visible, you just don't draw it, etc.

This is a fair point. All the mapped XWindows are certainly "retained" from this point of view.

But can you imagine something like Word being written without the "retained-mode" abstraction?

Yes, absolutely, and in fact I think it would be a much better program.