Case 1. when you say that you have files with 2/3/4 columns, does it means files with 2 columns have always same 2 columns, files with 3 columns have always same 3 columns and files with 4 columns have always same 4 columns?

Case 2. Is there any distinct identifier in the file name for these files with 2/3/4 columns?

If case 1 is true, then you can seggregate files with 2/3/4 columns in seperate directory & create 3 for each loop with one data flow task in each.

If case 2 is true, you can create 3 data flow tasks in single for each loop and then you can retrieve the file name in a variable and based on condition, you can move the control to load the file into one of the 3 data flow tasks..