Use your own java action. If the file follows a pattern you can implement that pretty easily with counters/modulus.
If it is a free form file, get basic csv parser and IMendixObject population going for the different sections, just modify the files for testing. Then modify your code it to test and detect your sections Thereafter the section detection and parsing/population code can be combined and tested again.
Alternatively write a parser that populates a domain representation of the file (file 1-* row 1-* field) and parse the rest in mendix then drop your abstract representation
This is interesting because the aggregate nature of the data contained in the csv
You can check out the microflow that generated this here
Have you tried adding and using the CSV Import module from the appstore? It lets you import and process csv files. You would probably have to cut your file into 3 files: every file with 1 header row and subsequent rows that follow these colums.
Bit off topic: Also have a look at the automated Excel file import now available in the sprintr. Button “Create app” and “Start from Data”