On this small scale test, this solution performs better.  I haven't had a chance to test on a larger scale.


SELECT *, pni.PONUMBER + REPLACE(CHAR(DENSE_RANK() OVER(PARTITION BY pni.PONUMBER ORDER BY PODATE) +
    CASE
        WHEN pn.PONumberSequence IS NULL THEN 63
        WHEN pn.PONumberSequence BETWEEN '0' AND '9' THEN 64
        ELSE ASCII(PONumberSequence)
    END), '@', '')
FROM #PONUMBERTESTIMPORT pni
OUTER APPLY
(
    SELECT RIGHT(MAX(PONUMBER), 1)
    FROM #PONUMBERTEST pn
    WHERE pn.PONUMBER LIKE pni.PONUMBER + '%'
) pn(PONumberSequence)


Drew