I would open the excel files in Filemaker, working with two native databases for the creation of the final database would likely be simplier.
Would all eleven million records be valid information or useful? Not empty or zero entries?
What is the purpose of merging this data into one table? It looks to me like it would make more sense and better data modeling to import the data into two different tables linked in a relationship, one with your 2656 GL Account records and one with your 4335 Cost Centre records.