Okay, take a look at the following code:

WITH SampleData (PERSON,TRANSACTDATE, STARTDATE, END_DATE, IN_PUNCH, OUT_PUNCH, [HOURS], PAYCODE) AS (

SELECT 1234,'08/03/2015','08/03/2015','08/03/2015', '06:00','09:00','3','REG1' UNION ALL

SELECT 1234,'08/03/2015','08/03/2015','08/03/2015', '09:00','13:00','4','REG2' UNION ALL

SELECT 1234,'08/04/2015','08/04/2015','08/04/2015', '09:00','13:00','4','LEAVE1' UNION ALL

SELECT 1234,'08/04/2015','08/04/2015','08/04/2015', '14:00','16:00','2','LEAVE2' UNION ALL

SELECT 1234,'08/05/2015','08/05/2015','08/05/2015', '08:00','09:00','1','LEAVE1' UNION ALL

SELECT 4553,'08/05/2015','08/05/2015','08/05/2015', '08:00','09:00','1','REG1' UNION ALL

SELECT 4553,'08/05/2015','08/05/2015','08/05/2015', '10:00','12:00','2','LEAVE2'

),

LEAVES_TOGETHER AS (

SELECT D.*

FROM SampleData AS D

INNER JOIN SampleData AS S

ON D.PERSON = S.PERSON

AND D.TRANSACTDATE = S.TRANSACTDATE

AND D.PAYCODE <> S.PAYCODE

AND S.PAYCODE IN ('LEAVE1', 'LEAVE2')

WHERE D.PAYCODE IN ('LEAVE1', 'LEAVE2')

)

SELECT SD.PERSON

, SD.TransactDate, SD.STARTDATE, SD.END_DATE

, MIN(SD.IN_PUNCH) OVER (PARTITION BY SD.Person, SD.TransactDate) AS IN_PUNCH

, MAX(SD.OUT_PUNCH) OVER (PARTITION BY SD.Person, SD.TransactDate) AS OUT_PUNCH

, SD.[HOURS]

, SD.PAYCODE

, RIGHT(LT.PAYCODE, 1) AS FLAG

FROM SampleData AS SD

LEFT OUTER JOIN LEAVES_TOGETHER AS LT

ON SD.PERSON = LT.PERSON

AND SD.TRANSACTDATE = LT.TRANSACTDATE

AND SD.STARTDATE = LT.STARTDATE

AND SD.END_DATE = LT.END_DATE

AND SD.[HOURS] = LT.[HOURS]

AND SD.IN_PUNCH = LT.IN_PUNCH

AND SD.OUT_PUNCH = LT.OUT_PUNCH

ORDER BY SD.PERSON, SD.TRANSACTDATE, SD.PAYCODE;

I'm not sure how this will perform, as a look at the execution plan shows that 98% of the cost is in two Sort operators. This may mean that a properly indexed source table would allow this to fly, but only testing can determine that. Sizable volume could also slow it down. Let me know...