How is the CPM calculated in Radlog to get the decimal?

As the title suggests, I'm curious as to how the CPM calculated in Radlog to get a result with a decimal. In all of the counters I have programmed, or played about with, I have always had whole numbers. Generally my method is to fill up an array of 60 indices over 60 seconds, add them together and that gives my CPM. Then each second overwrite the oldest. Its like a conveyor belt! This is the way the Net-IO GC10 calculates the CPM. For faster sampling simply reduce the array to 30 or 15 indices and 30 or 15 seconds subsequently and multiply it by 2 or 4. All iterations of my method of calculating CPM always, always, result in a whole number. So how is it done in Radlog? I can only assume some kind of averaging. But what is it?

 

The incoming readings sometimes can have different sample periods for some counters. e.g. 10s or 30s depending of CPM count.
Sometimes the readings become unsynchronised, sometimes counters cannot be configured for a different data submission rate.
What radlog does is to add up the readings that arrive during the submission period and average them over time (a bit like a histogram?) so that the CPM value is real for the submission period. So that information is not lost, the calculated rate is non-integer.
When data is checked by performing an average, the average value should match.
If the calculated CPM is rounded, this match becomes less accurate.
I checked the calculations for accuracy when I wrote Radlog and I think at least two others have check their own measured data and what is displayed on radmon.org, and found the numbers to agree (thankfully )

Going over some older topics I thought I had replied to this. I did say thanks, no didn't reply.

Thanks Dan, that makes perfect sense.