I mean a real continuous data scatter plot. See pages 41, 42 and 43 of my MSA presentation. The scatter diagram is on page 43.
The plots you did are Multi-Vari Plots (as described in the presentation)
The first 3 with the spec limits are the correct plots. You need to see the repeated measurement variation vs. the part to part variation vs. the tolerances. this is exactly what all of the mathematical equations are trying to 'quantify'. Removing the spec limits might make you see the dots 'better' but that isn't the point. In fact perfect (no) measurement error would have the points fall on top of each other. It also distorts the operator to operator differences as the y axes are no longer the same size (a basic violation of graphical display rules). So removing the spec limits doesn't clarify anything - it muddies the view.
Miner plotted scatter diagrams. and his axes are equivalent for each operator. He is missing the spec limits which artificially 'magnifies' the variation of the repeatability and the part to part variation. But this is a good first pass...
One other point is that the parts selected for the study do not vary much - an MSA should have parts that vary across the natural range pf variation. perhaps your process is that tight but maybe not. The math would say that that "%PV" ratio would be quite bad and would indicate that you can't really apply SPC...