Exactly, I would do the same thing.
First, split each watershed into parts by municipio and put the result into a temporary table:
SELECT * INTO [watersheds_by_municipios]
FROM CALL GeomOverlayTopologyIdentityPar([Watersheds_Hu12], [Municipios],
Then select just those watersheds which have a part that is 75% of the total area or bigger:
FROM [Watersheds_Hu12] AS w INNER JOIN [watersheds_by_municipios] AS wm
WHERE GeomArea(wm.[s_Geom (I)], 0) / GeomArea(w.[Geom (I)], 0) * 100 >= 75;
The result table of the second SELECT has no unique indexes, but we know from the definition of the task that each watershed can only participate once (there can be only one part with 75% or more area). So, we can, for example, alter this second SELECT to do SELECT INTO and then add an index on the MFD_ID field manually (Add Identity button in the Schema dialog should do the trick).