The reported CPU usage is much higher or lower than top reported value?

1) testcase: gops cpu usage >> top cpu usage
in my testcase, there're heavy RPC request/response, create goroutines, chan read and chan send.
The gops reported CPU usage is much higher than top reported value. What's the differences btw the methods used by gops and top? Which one is more accurate?

2) testcase: gops cpu usage << top cpu usage
in this testcase, there's some sync.Mutex lock and lock, chan send, chan read.

I learned gops uses go runtime statistics to calucuate it, and top uses syscall to calculate it. If I care more about actual system load, top is more accurate?