Issue when joining 3 temp tables - sql-server

Question

I have an issue joining 3 temp tables.
I am going to write the whole code but the thing that doesn't work is a join when selecting all three temp tables. Everything else works just fine (type just in case if someone wants to see the whole picture).
I need to join all three tables on dates to make sure I select the same inventory during the same time period. Whenever I join those three tables I either get Forecast or Actuals right, but never both.

Answer 1

When I type this I get Forecast correct, but Actuals incorrect a.[DMDPostDate]=u.[STARTDATE] and f.[STARTDATE]=a.[DMDPostDate] (Forecast correct 6998.649, Actuals are not correct 826)
-- AND u.[STARTDATE]=f.[STARTDATE] and f.[STARTDATE]=u.[STARTDATE] (Actuals are correct 10369, Forecast not correct 8322.315)
-- and a.[DMDPostDate]=f.[STARTDATE] (Forecast correct 6998.649, Actuals not correct)
-- AND u.[STARTDATE]=a.[DMDPostDate] (Forecast correct 6998.649, Actuals not correct)
-- AND u.[STARTDATE]=f.[STARTDATE] (Actuals are correct 10369, Forecast not correct)
-- and u.[STARTDATE]=f.[STARTDATE] and u.[STARTDATE] = a.[DMDPostDate] (Forecast correct 6998.649, Actuals not correct)
From your post - it seems fairly clear that joining STARTDATEs gets you the correct actuals, and joining DMDPostDate to either start date gets the right forecast.
Something thing to consider - u.[STARTDATE]=f.[STARTDATE] - this clause should have no impact to the A table join, what if you added this to the F table ON clause, and then u.[STARTDATE] = a.[DMDPostDate] to the A table clause
The way it's working now, you're left joining everything from table F to everything it can join to in table A based on the U.[UPC]=F.[DMDUNIT] AND U.[MASTERCHAINNAME]=F.[LOC], without any dates. It could be a challenge with the way it's evaluating the clauses as a result.
Dark horse answer - you don't need to join on any dates - actuals are correct when you aren't joining dates for the A table - when you use AND u.[STARTDATE]=f.[STARTDATE] it may be applying to the U/F join which could lower the forecast number.
You could also troubleshoot by joining the F and A tables separately to the U table to see if you're getting the expected values then.
Best guess -
SELECT
U.[UPC] AS 'Item',
U.[MASTERCHAINNAME] AS 'Chain',
U.[STARTDATE] AS 'Start Date',
U.[EVENT_TYPE] ,
U.[EVENT_NAME],
SUM(F.Forecast) AS 'Forecast',
SUM(A.HistoryQuantity) AS 'Actuals'
FROM
UDT_CKB_SNAPSHOT U
LEFT OUTER JOIN
FCSTPERFSTATIC F ON U.[UPC] = F.[DMDUNIT]
AND U.[MASTERCHAINNAME] = F.[LOC]
AND f.[STARTDATE] = u.[STARTDATE]
LEFT OUTER JOIN
HISTWIDE_CHAIN A ON U.[UPC] = a.[DMDUNIT]
AND U.[MASTERCHAINNAME] = a.[LOC]
AND a.[DMDPostDate] = u.[STARTDATE]
GROUP BY
U.[UPC], U.[MASTERCHAINNAME], U.[STARTDATE], U.[EVENT_TYPE] , U.[EVENT_NAME]