update statement with FROM and ctids

I have an update statement of the following (simplified) form:

     UPDATE changes.table1 a
        SET checked_out_at = a.checked_out_at
       FROM unified.view2 b
      WHERE a.ctid = '(4,48)'
        AND b.some_pkey = a.some_pkey
  RETURNING *;

This statement will just fail to update. I see 0 rows and UPDATE 0 when I run it. But if I modify it by dropping the joined table, the update will go through (just once, the ctid changes after all, and I'd need to find the new one):

     UPDATE changes.table1 a
        SET checked_out_at = a.checked_out_at
      WHERE a.ctid = '(4,48)'
  RETURNING *;

Unfortunately, I need the joined table for some of the data in the RETURNING clause.

Why are these two features incompatible? Do the joined rows somehow have ctids different from the single table? I know that I don't have a proper understanding of all the low-level system columns, but this still seems rather straight forward. Can this or something similar be made to work?

If the goal is to limit the number of rows that get updated, consider using a subquery as follows:

UPDATE changes.table1 a
   SET checked_out_at = a.checked_out_at
  FROM (SELECT c.ctid AS original_ctid, b.*
          FROM changes.table1 c
          JOIN unified.view2 b
            ON b.some_pkey = a.some_pkey
         LIMIT 10) d
 WHERE a.ctid = d.original_ctid
RETURNING *;

This will limit the number of updated rows while avoiding an extra table or index scan since a TID scan will be used to find the rows to be updated. It also eliminates the need to determine CTIDs in separate queries.