Data after retention period is not correct

The Unbounded Analytics is operating on the data-set that's in the data-store for the selected timeframe. When the timeframe is outside of the (full-granularity) short term window that is defined in the self-hosted configuration the results are imprecise. In that case we display an indicator in UA and/or the timeframe selector to make clear that the results are not precise. This means that the saved data is no longer representing the real data and while we state we save 1% of data this is 1% of the whole data and not 1% of the data per service.

There are multiple approaches how to continue:

increase the data retention period 
so that it covers the period you want to analyze. Please be aware that this will increase the utilization of the compute and storage resources. 
=> Knowing your OnPrem physical hardware resources are exhaused this is a no-go...
create daily API calls to retrieve the real numbers
reading the explanation above and following our joint analysis we found and concluded that the data not in retention timeframe is very close to the real number. It would therefore be an option to run an API call retrieving the count allowing you to compare this number for covering the use-case
enhance the Instana product for this use-case
while from customer perspective this will be the most desired one the complexity of building this into the Instana architecture will be huge... Reason being that all calls are tagged with data not resulting in a typical timeseries metric data we want to maintain but a highly complex multi-dimensional dataset needing high velocity calculation and costing more resources and performance for realisation...
=> We will keep this item on our radar and re-evaluate options during 2022 when our own histogram based new metric store is further adopted into the product looking at options if this can address the use-case

For now we advice to consider option 2 until further notice. In the meanwhile we will be moving this ticket to a feature request for future tracking.