remove-package --source-only doesn't do as advertised

Earlier today, I did the following:

(base)adconrad@cthulhu:~/utouch$ remove-package -m "Renamed to evemu" -S utouch-evemu
Removing packages:
 utouch-evemu 1.0.9-0ubuntu1 in quantal
Comment: Renamed to evemu
Remove [y|N]? y
1 package successfully removed.
(base)adconrad@cthulhu:~/utouch$ remove-package -m "Renamed to frame" -S utouch-frame
Removing packages:
 utouch-frame 2.2.3-0ubuntu1 in quantal
Comment: Renamed to frame
Remove [y|N]? y
1 package successfully removed.
(base)adconrad@cthulhu:~/utouch$ remove-package -m "Renamed to geis" -S utouch-geis
Removing packages:
 utouch-geis 2.2.10-0ubuntu1 in quantal
Comment: Renamed to geis
Remove [y|N]? y
1 package successfully removed.
(base)adconrad@cthulhu:~/utouch$ remove-package -m "Renamed to grail" -S utouch-grail
Removing packages:
 utouch-grail 3.0.5-0ubuntu1 in quantal
Comment: Renamed to grail
Remove [y|N]? y

I had assumed, both from how the previous remove-package.py worked, as well as from the --help text, that this would leave the old binaries in the archive, so we could transition smoothly and then reap them from the NBS list.

Unfortunately, the API seems to have different ideas about this, and forces the deletion of binary children:

14:06 < Laney> hah, comment in LP's source:
14:06 < Laney> # Special deletion method for the api that makes sure binaries
14:06 < Laney> # get deleted too.

Filing this as an ubuntu-archive-tools bug for now, as maybe it's just a question of not trying to do that, though this functionality is actually pretty handy for doing soft transitions by intentionally orphaning binaries so they show up on reports, so perhaps it's ultimately an LP bug to extend/fix the API as well.