I'd have to rewatch the video, but I'm pretty sure one of his specimens was new oil, poured from the bottle into an empty engine. Then he took a sample from the sump, without running the motor, and did a particle count on it. Additional counts were done after running the engine. I think some were done where he purposefully introduced particulate contaminant to see how much the filter would/could remove.
The conclusion was that merely pouring the oil into the test stand motor made the oil magnitudes "dirtier" than it was from the bottle.
Granted, his videos are pretty 101 compared to bitog tribal knowledge.