We are able to parse a variable length file by setting "is_record_sequence" to "true". However the parsed data is "shifted" left by 4 bytes. We do have 4 byte RDW headers in the data. It seems Cobrix parser assumes that in addition to 4-byte RDW header there is 4 byte BDW header present in the data and hence consumes first 8 bytes from each row of the ebcdic data file effectively shifting actual data left by 4 bytes.