I did this recently in PostgreSQL. Are you in a position to use a temporary table? If so, for each duplicate set, insert the MIN() primary key into your temp table, and then do your UPDATE using a where clause using the PKs in the temp table.

Edit: following your comment, here is something I did recently.

CREATE TEMPORARY TABLE misc_updates (

temp_oid INTEGER,

job INTEGER,

run CHARACTER VARYING(8),

quantity INTEGER

);

INSERT INTO misc_updates (temp_oid, job, run, quantity)

SELECT

MAX(runjob.oid) temp_oid, runjob.job, runjob.run, SUM(runjob.quantity) sum_quantity

FROM

runjob

INNER JOIN job ON (runjob.job = job.code)

INNER JOIN

(SELECT run, job FROM runjob GROUP BY run, job HAVING COUNT(*) > 1) my_inner

ON (runjob.run = my_inner.run AND runjob.job = my_inner.job)

GROUP BY

runjob.job, runjob.run, job.quantity

;

/* Do updates on one of the duplicated runjob rows */

UPDATE runjob

SET quantity = mu.quantity

FROM

misc_updates mu

WHERE

runjob.oid = mu.temp_oid;

You can swap 'oid' for your primary key (my problem was that the table had no primary key!). Also the critical thing is the where clause in the UPDATE, so only some rows are updated. You'll need to swap out MAX() for MIN(), and of course change the rows to the ones in your use-case. Bear in mind that this is for PostgreSQL, but the approach should be much the same.