Performance Manager - CPU usage on Linux hosts

Hi,

We've been relying on WLA/PerfMan for years to accurately see the individual CPU metrics for RedHat hosts on a variety of boxes. Our UNIX admins also use a tool (previously BrigBrother now Xymon I believe). Anyhow, you know how most 'nix admins rely in load or load average for determining such "good/bad" analysis, well I've noticed as of late that ECC does not seem to be tracking thread or CPU utilization accurately and wonder if others have seen this as well. Probably requires a case and the typical gathering of logs/grabs but maybe this is explainable by others. Basically I have documented how if I rely on ECC's metrics of host system % proc usage, %idle, or any of the "%" metrics (sys, user, etc) that even if the host has multiple stuck procs generating a load of say 4 on the box and I mean 100% CPU time for days, in performance manager, the % util bounces around from 30-60%, in fact I cannot find a single instance of 100% util even though I am positive there should be at LEAST 4 threads reflecting the 100% time. I don't get it. This cannot be due to averaging out because 24hrs of 100% util should show those threads at 100% all day - right? Now, on windows boxes - no problem. 100% means 100%.

Anybody have any ideas or is this a bug? I lean towards bug.

thanks

Dave