The “cloning” procedure mentioned above is oriented at an environment where all AM nodes have exactly the same configuration (cattle – not pets). A single install is done, and each AM node points at the same configuration store. There are restrictions with that model noted above (see notes on sticky load balancing, etc.)
If you want to have two different AM configurations (primary / secondary) you should be able to script that using amster. You will have two different amster imports, or one import, and apply some kind of search/replace on the elements that you want to customize for each install.
Tl;DR – Everything that your ssoadm script does you should be able to replicate using amster.