One Star

Can't get tExtractPositionalFields to work as expected

Hi all
I have a positional input file that has multiple record types (different length, different fields) in it. I need to process this into different database tables (each record type goes into a different table).
I have a working solution as follows:
* read in all the records
* split with a tMap according to record type (first 4 characters of each record) and put outputs into different temporary files
* for each file have a subjob that reads the file with the correct positional schema (all coded as metadata "Generic Schema")
* map each file with a tMap onto the corresponding database table
So far so good. Now I thought: why not do this all in one step? I started to try by using the same input file, the same tMap to split record types, and then use a tExtractPositionalFields with the same Generic Schema that I used for reading the temporary file above. The picture gives an idea how I wanted to do this.
The problem is that tExtractPositionalFields does not seem to use the generic schema. Rather it always uses the value in the "Pattern" field, although I did not activate the "Customize" checkbox. I can get it to work when I put in the right pattern string in the tExtractPositionalFields. But this is not really as desired, because it then detaches from the Generic Schema, and changes in the future will have to be done twice. Also if I put "" or an empty string into the pattern textpox, the corresponding error is thrown from the Java code.
Am I understanding the tExtractPositionalFields component wrong? Or is this anyway stupid what I'm trying to do and I should stick with the temporary files?
Thanks!
CU, Joe