On Linux systems load averages cover different resource types (not just CPU, but also other resources like disk, uninterruptible locks), so you cannot just divide by the CPU count. 

More info on this topic can be found here: http://www.brendangregg.com/blog/2017-08-08/linux-load-averages.html

Thus, by default, in Linux template we should use "system.cpu.load[*all*,avg1]" instead of "system.cpu.load[percpu,avg1]" and so on...