I may be way off, but i believe the goal is to "slice" off 98 bytes of data in the 256 lrecl file and then start the next 98 byte output starting with pos 99. When the "end" of a 256-byte record is reached, the last 46 bytes should start the next 98-byte record and the first bytes of the next 256-byte record should be appended so there are no breaks between byte 256 of the first record and byte 1 of the second record. . . Whew. . .
As i mentioned, i may be way off, but looking at the abbreviated sample data. . . .
Hope this helps,