I'm not sure this is a FOP problem, or if it is the Docbook stylesheets that is the problem. I suspect that FOP would need to do the work anyways.

The image callout tag in Docbook (any version) is not useful. When writing user manuals for software, a not insignificant part of the work is creating screenshots and marking parts of the screen. If the screens could be "tagged" with callouts in a structured manner such as Docbook, then the workload would be significantly reduced. People use all sorts of horrible solutions such as doing it in Word, Powerpoint, or many different programs. 

One authoritative source has a work-intensive workaround. 
http://norman.walsh.name/2006/06/10/imageobjectco

This documents the tag:
http://www.sagehill.net/docbookxsl/Imagemaps.html

Current status with FOP:
Callouts are not rendered on top of images
Callouts are rendered in a calloutlist, below the figure (using standard stylesheets).

Expected behavior:
Given the following input:
    <figure>
      <title>The main screen</title>
      <mediaobject>
        <imageobjectco>
          <areaspec units="calspair">
            <area coords="0, 0 1000, 1000" xml:id="firstcallout"/>
          </areaspec>
          <imageobject>
            <imagedata fileref="screenshots/mainscreen.jpg"/>
          </imageobject>
          <calloutlist>
            <callout arearefs="firstcallout">
              <para>A first callout</para>
            </callout>
          </calloutlist>
        </imageobjectco>
      </mediaobject>
    </figure>

FOP should ideally produce a PDF file with a figure looks like "mainscreen.jpg" but is actually "mainscreen.jpg" merged with a "1.svg" file at the coordinates (callout number 1).