Funny how we see things differently Sven. Admittedly the R squared values aren't much to write home about, but if you plot each batch separately then three of them clearly show velocity falling as pellet weight increases, and one batch is far more random and gives a trend line that is almost horizontal, if anything rising rather than falling. And that is only due to one particular reading, 759 fps and 8.48 grains, and that might even be due to a small error in data transfer.
I notice also that all of the pellet weights are given to the nearest 0.02 grain; perhaps the scales used had this as the resolution. As we are talking about a pretty small absolute range of pellet weight values, then a coarse resolution is bound to open up the precision.
Maybe I should have used a less weighty adjective than strong; the result is certainly clear. Very crudely, pellet weight change from 8.3 to 8.4 drops the velocity by something like 2 fps.