They used an EOS R, not the RP, for the tests.(read the fine print) No big deal.
Well... What we did was a so called "linking measurement" (or "associative measurement"; I hope one of them is a proper translation for what I mean), quite common in science and industry to e.g. prove batch confirmity.
The idea in this special case: Neither Canon nor any other camera manufacturer do actually develop their own memory card controller or at least IP-logic. Also image processors usually don't change that often. It's often the case that the same silicon and firmware is used across various cameras and therefore real world performance don't differ either. For instance: There's no significant difference, in terms of memory card performance, between a Sony A6000, A6300, A7 II, A7R II and some more models released in this period.
Instead of measuring the same data over and over again, we prove that camera A has the same characteristics as camera B. It took some expierence, but usually it's easy to guess which hardware is used to have a dataset to begin with (EOS RP: It's Canons second UHS-II consumer camera. Not that hard^^). As mentioned in the notes, we did a 6-point examination for the EOS RP. We always use the fastest card from the reference dataset (for obvious reasons), if available, one card with a special characteristic ("card x" performs way differently in camera A compared to other cameras) and some randomly picked samples.
If the difference between camera A and B is about the same as measurement uncertainty, it is OK to estimate that other results will match as well. What is important to us: The (small) differences between camera A and B must be randomly as well, not having a trend like "it's always ~1.x percent better". Otherwise we do the complete test, like we did with the
Nikon D5600, even if the difference compared to a
D5500 is not that big at all.
In case of the EOS RP, it took us some minutes to find the correct settings at first (to match the speed of the fastest card), as the Canon EOS R/RP's writing speed depends heavily on file size, as explained on our website. Afterwards, all samples were below 2% difference.
Finally: No big deal? As explained (hopefully), it's not like looking in the crystal ball and saying "yeah... I guess we can use the EOS R data". And you have to do hours of measurements for the EOS R as well
Interesting but, it didn't include two very fast cards. Pro Grade has [...]
...unfortunately not replied to our sample request(s). We'd really like to review their products.