Collocate in Graph Builder does not allow to select the master file. It always takes the file in the first read module as the master and the file in the second read module as the slave.
The GPT file of the graph can be edited selecting a master file under the masterProductName in the GPT file. However, SNAP Batch Processing will always select the file in the first read as master, regardless what is written in the masterProductName.
When running Batch Processing it only allows to select first input files, meaning that it only allows to select master files, taking the slave from the second read module of the graph xml file.
This is nonsense: Collate in Batch Processing only allows to collocate many masters with a slave, whereas the usual processing is to collocate many slaves with a master.
There is a workaround. It is using ProductSet-Reader instead of read for the slave files, adding the slave’s files that need to be process in the ProductSet-Reader list. Very frustrating, but it works.