Zooming in on plot results in visual change in binary Calculation from always 0 to sometimes 1 . Bug?

I implemented a PI Vision Calculation to implement a binary trend that shows a '1' spike when a value suddenly changes. My goal was to identify at a glance sudden changes for long x vs time plot time durations (those events happen to be very rare). The behavior I expect is that zooming will not change the trend. The behavior that I get is that if I use a 1 day or 1 month duration, the Calculation seems guaranteed to display as '0' on plots (I assume because the '1's are rare); however, when I zoom in, the calculation may show up as '1' during the reduced time duration. I confirmed that zooming back out makes the '1' go away again. The behavior gives me the idea that some trend data points are filtered for performance.

As for my business case details, I don't have the bandwidth to obtain and learn AF, nor do I have the influence to get an AF developer on this project; we can't afford to miss the events, and if I need a <=15[minute] plot duration, which would result in >=96 clicks per day per screen (10-20 screens), I will use a different tool to process the data offline like Python or MATLAB. For my application, I need to identify at a glance all times the binary trigger should go to '1' using 2[s] data resolution for a 1 week duration of data (I've assumed I'd use a value vs time plot). If it is indeed a performance optimization at the root of the behavior, for my application 90% of the 0's can be filtered out, but I need all the 1's in the plot.

Thank you for your time.