Dashboards with large timespans reduce the peaks and are less use-full

The "problem" I have is if you take a look as an example to the timings of the garbage-collection graph.
I have here an app who does every 5 seconds a 300ms gc, if you are in the 10 minutes frame this is visible. I can go back 24 hours in this granularity. But if it goes out of this 24 hours frame, the gc time reduced to 22ms(resolution 1 minute if it sees correct).
I know where it comes from, but this makes the graph useless, as it told me the gc time is 22 ms and not 300ms.

Similar happens, if I switch from 10 minutes frame to 30 Minutes frame. The max time goes down to below 150ms, which may also leads to wrong decisions. Larger peak in the 10 Minutes(1s rollup) view by 16 seconds in the 30 Minutes(5s rollup) frame 3.2 seconds. I was here in the live view.

If something happens in the last 24 hours, i can dive deep enough to identify the exacter timings. But if it goes out of this, I can't see the real values.

Normally i take a look at the last 24 hours, and if there something special, I make a deeper look. But if the graph shows me 22 ms i don't will do it. This is for every graph more or less and goes more invisible if you have larger deltas between max and min values.

Maybe something like save for every rollup the max value. So we have for the 1 minute resolution a mean value of 22ms but the peak is at 300ms, and so on, and also the possibility to visualize it in the graphs.

Hope this is now not too confusing. The problem I think are both points.