On my new 10.5 implementation, a CSV Windows Machine Asset import is resulting in the system connecting to the RMS Information Server, launching a bvJobProcessor and running queries on the network (as evidenced by new QE stub processes being launched).
Everything needed for the import is in the CSV file it is configured to read in the Data Collector configuration, so why should it use the default data collector as well? It is not configured to use the Default data collector at all.
If I shut down the Information Server service, it does read things from the CSV file, but gives an error about not being able to access the Information Server. In addition, it fails to import the TCPIP address data for the assets, even though they exist in the CSV file.
This doesn't happen on my legacy v9.01 system. On v9.01 it collects the assets from the CSV file, all records successfully, and nothing more - with no errors - as it should.