First of all it is not a sample. It is the entire population stratified. You can't make any type of determination because your data is biased. There is something about your data in general that is offensive. It could something as simple as your chronograph causing the bias. Something is prejudicing the data. There should not be that much variance in your standard deviation based on 1/10 of a grain. In each stratum though your standard deviation is too high to be acceptable. Example, column D is standing proud and dominating the rest of the strata and column C is close to what one would expect from a match factory load. You're going to have to figure out the bias and get a bigger population. Those numbers don't correlate at all with what is expected when all things are relevant; i.e. loads are accurate, chronograph is accurate, weapon system is sound, the group in the targets, and so forth.
As for your question above? Take an average from all your data sources to decide on a beginning load after checking the brass from that population for signs or lack of signs of excessive load. Send me a PM with your exact load data so I can research it a little.
Also, you don't need that energy row if you are taking that data from a loading manual and just plugging it in. If so, it is not doing anything in the spreadsheet. I'm assuming it is not a measured average for each 5 shot group.