In order to make the charts responsive, they are limited to 250 data points at a time. As you've found, the data will be loaded as folks zoom in to a tighter time range. There is no current workaround in VMAN proper (barring the integration with Orion).
What is the use case for users to get that granular of information in a monthly report? Other than "they just want it"? I'm digging to understand, because this is the first time I've heard this requested. As you mentioned, the graphs become almost unreadable at that level with the raw data. We did add a "force Raw" reporting option to the latest version of Storage Manager, but the use case there was to troubleshoot a specific issue that occurred. So theoretically, you would be interacting with the charts as you are now by zooming in to a level. You could also report on the monthly value peak, or weekly peaks, if you wanted to see very high usage periods.
Finally, there are no 12-hour roll-ups. There are the hourly or 24 hour roll-ups on the hourly scale.
You are 100% correct on the 12-hour part - I was bouncing between a lot of things and my brain pulled the 12-12 markers and ran with a 12-hour increment.
The use-case for this is that we have a group that is very intent on keeping an eye on their peaks (mostly CPU related) on a monthly reporting basis. They have high workloads every few evenings that we know cause peaks, but they have no desire (currently) for weekly reporting round-ups. Just monthly. With the 24-hour normalizations in VMAN graphing, we aren't able to reliably show the workload their systems are under.
We have tried using the peak-usage metrics, but went through a long support back-and-forth because the values were not trusthworthy at all (IE: reporting >100% CPU usage).
For the time being, and the foreseeable future we are using the monthly performance reports directly within vCenter for the specific systems.
But my $0.02 is that providing reporting out to groups/customers/etc on a monthly (as opposed to weekly) basis would not be that uncommon. With the amount of normalization occurring in VMAN makes it more difficult to provide graphs that more closely portray a specific server's load when pushed past 1 week. Not to say that I don't understand where it is coming from in terms of readability, but sometimes you don't need to read a graph in finite terms to see the trends you need to see.
Ok - so the CPU Used % counter that had problems should be replaced by either CPU Utilization of Host % or CPU Load. Unfortunately, it doesn't appear that either of those have a "Peak" attribute equivalent, so that will need to get addressed on our end. Once that's done, you should be able to generate these same reports for those teams and see the peaks for the time period in question.