It is expected for sampling to happen when you add a secondary dimension.
Unsampled aggregate report vs Ad-hoc report subject to sampling
GA creates a set of unsampled, pre-aggregated data tables, which are processed on a daily basis. These “aggregated tables” are used to quickly display unsampled reports that you see in standard reports.
When we modify the standard report such as by adding a secondary dimension, GA ceases to pull data from the unsampled aggregate table. Instead, GA now has to do an “ad-hoc report” and query from the raw data instead of GA simply looking at the readily available aggregate tables. This causes sampling to occur.
Take note that “Page” is a dimension with a large number of unique values (also called cardinality). A high cardinality dimension in an ad-hoc report means there are more data to be queried.
As a workaround, I recommend the following options:
1. Shorten report date range. I checked the date range for Dec 1 -15 and Dec 16 - 31 and sampling did not occur.
2. Create a Custom Table
2. Download Unsampled Report
Google has 14 months threshold for higher sampling limit. Previously, it was 13 months but a feature request has been raised by other clients which were pushed to production.
Currently, 100 Million is the higher sampling limit, but if the data is older than 14 months then the max sampling limit is 1 Million.
The unsampled report, however, can be requested for any date range (even before the property is not yet converted to 360).
When someone purchases GA360 license, the view will get a 14-month back-fill of sampling data approximately 48 hours after becoming 360.
If the date range the client is using hits the sampling thresholds mentioned in this Help Center article: https://support.google.com/analytics/answer/2637192?hl=en#thresholds, the report will be sampled.