We did a fishbone on the method and came up with 3% variability… assuming your analyst is not a shit head. Youd only weigh once and then its just volumetric. Scale, piston pipette, air cushion pipette dilutions. You can actually measure all variables at each step independently, or trust the vendors for some. So uncertainty is “certain”… lil jk there. But truely I find all that BS, it is all theoretical, practice is where you get the real story.
So you put that particular method in practice with some homogenious hemp sample (or maybe real life customer duplicate samples), you do this say 20x. Then you get the real story on the duplicability of your method. That particular method being 5% on homogenious material and a little less than 7% on real life samples.
But I dont not like the method you just laid out. The double extraction is a nice thought. Just it is quite impractical at large scale and unnecessarily complicated with too many quantitative transfers. You are not getting all the extraction solvent (thus analyte) when you transfer and filter.
Edit to say: you could use an IS in the extraction solvent to control for losses in your method id guess, but my state dont allow those kind of controls.