I don’t know about you but when I am using FSD it is pretty clear I am driving.
In any case they are completely different, even if you stipulate that Tesla has a safety driver - as you so perspicaciously point out, there is just no way to compare.
The point had to do with the data collection by Waymo and Tesla. In either vehicle, when the car is actively operating the controls, the data generated by the car is just as useful to improving the software as if you were sitting in the rear seat (I don't recommend doing this in a Tesla!). In fact, even if you are manually driving, if you have FSD, the car may be running the FSD planning and generating data back to Tesla based on what it would have done versus what you did. So, even if you are in full control of the car, the miles driven may be useful to Tesla.
The other point is that both companies are still in active development of their AV products and uses the products in the field primarily as data generators for those development efforts. There are some secondary goals. Waymo is obviously using their product to introduce the public to driverless vehicles to accustome them to the concept. Tesla rolls out FSD beta to it's customers to help meet their obligations with FSD. But, I believe that the primary purpose for both companies (as well as Cruise), is to generate driving data to support their developments.
So, from an engineering perspective, a mile driven under self-driving of any type, is equivalent, assuming, of course, that appropriate data is collected! From that perspective, Tesla has the potential to be collect orders of magnitude more data than Waymo or Cruise.
If you are using miles driven by each type of vehicle as a measure of goodness, then no, the two are not comparable. That's been argued here over and over.