> I wonder if there is any other solution I have no clue about.
Not really. FM can't Xref to anything but Xref markers, which can only be placed in text flows (not even in graphics text).
When people want to hyperlink to a graphic, they use text frames, either in front of or behind the graphic (and behind usually implies some confidence that the graphic is opaque in both preview and final rendering).
If I wanted the target of a caption Xref to be the graphic, I'd put the caption (or a clone of it) on top of the graphic using a color set to invisible by Color Views at render time. Give it a unique para tag, such as Caption.Invisible
The graphic's invisible text frame could contain just:
- a cross-reference, by paragraph text, to the real caption below the art
- an Xref Marker, auto-created when you create the See/Refer-to reference to the Caption.Invisible in the adjoining narrative
Any changes to the real caption are made to the always-visible text below, and are picked up automatically by the invisible clone.
Extra work to set up. Trivial to maintain. Which is to say, typical FM.