Aggregated results too high when using multiple servers via "client" command

I am trying to do a benchmark from multiple clients in order to test the performance of a NAS device.
When running the test each of the clients reports a performance of ~950MB/s. The summarized performance in the log shows 4x950MB/s though the NAS device only supports (and shows) a maximum bandwith of 950 MB/s in total. To my understanding, the servers I am adressing via „client=host.list“ should generate load in parallel which should then reduce each stream to the maximum bandwith supported by the NAS divided by the number of clients… but somehow each client reports the maximum bandwith oft he target and the summary therfor is wrong. I tried group_reporting and various other command without success.
Any advice here for a newbie?