It’s partially how the sample is presented / handled. They’ll separate the sample into a couple of different lots and analyze whatever it is that they’ve received. I believe flowers (for potency) are first desiccated to some nominal water content prior to weighing. Terpene analysis is likely refrigerated until needed or immediately placed into solvent. I’ve been meaning to have a conversation with the lab director on some of their procedures.
When you look at the photos of the samples as presented to the lab, a good job trimming does appear to make a difference as you’ve removed the mass that doesn’t contribute much to the percentage of cannabinoids (compared to the total mass). So, what you’d probably want to hand them is basically the most tricome laden portions of the flower. You’re trying to not give them seed laden buds or a bunch of stem and certainly not the whole plant.
A good deal of the “error” could occur here as it entirely depends on what you give them lot to lot. Seeds, stems, leaves, etc could have a significant negative effect on the analysis. You could also sprinkle on some kief or add some synthetic terpenoids to “cook” the sample. I don’t know if anyone does this, though.
So when we say 25%, we are talking about the juiciest part of the plant. The well trimmed and handled flowers do show differences but the photos of the low percentage cannabinoid samples illustrate the bulk of the poorly handled/trimmed samples as submitted.
Something like this:
will probably test higher than something like this: