Your first option is not a good one. Generic key reads require you start at the first byte of the key. Since you are starting from the sixth byte of the key, you would have to use READ NEXT to read every record in the file to select the records you want.
Alternate index (AIX) may -- and that depends upon your data -- work. The key relevant limit is that an AIX with duplicates is limited to 32767 occurrences. If any of your PART2 values occurs more than 32767 times, you cannot use alternate index -- in which case your only choice will be a complete redesign. http://www.redbooks.ibm.com/abstracts/sg246105.html?Open
is the VSAM Demystified
redbook and a good place to start. Also read the DEFINE ALTERNATEINDEX section in the Access Method Services for Catalogs