I would punt, but if you are forced...
8133 is really slow stuff, so just be aware how far down on the charts you are going with that.
You are correct that the steps are too coarse and too few to call anything without lots of risk that the next time you shoot you would end up with more confusion.
There is no point to an SD when you have only three shots. Even with five to seven shots, it is risky to make a claim on an SD, so we look harder at the ES. A good one here prolongs the testing, but a bad one cuts that charge loose.
At least some of those are fair for the top end of your test, but with only three shots per step, this is very risky and forces you to take more shots there to see if the trend holds or falls apart.
I would either punt completely and trade for some H4350, or if forced to continue with 8133 I would explore those top two groups and see if a larger sample plays out. Maybe drop a point in between 66 and 66.5 to see if there is any hope for stability there, and then tune the seating depth.
Your odds at this point are only about 50/50 that the test shows anything at all. The 66.0 and 66.5 groups could actually be about the same size if you shoot more samples. The only saving grace is that you might then be able to play with seating to tune the group down since the velocity looks stable.
ETA: I would also suggest you run any follow-up tests at distance. If the steps from 66, 66.25, 66.5 show a POI node with your current jump 20, then the follow up with seating depth has hope.