Just been through this and SSIS is very particular when reading from Excel. It samples the 1st 8 rows and determines datatypes from that sample. As we all know, it is tricky changing data types in SSIS so I found it easier to force the datatypes to be correct in the originating excel. So, rather than hope my datefield column happened to have more than 4 date values (if say the sample had 5 nulls, SSIS would assume string) I faked up the 1st 8 rows. In this way, I could fill out the fields to ensure the data types were right and truncation was avoided.

Remember, 1st 8 rows are sampled, easy to fake up