First, four pins is an over constrained solution to this problem. Only two are needed to locate the top plate and the most sensible arrangement would be a central hole that the top plate pivots on and a single pin located as far from that central hole as possible in the bottom plate with a second pin there. The top plate would have five holes, one in the center and four for the 90 degree rotations.
Now, which method would be the most accurate? The answer to that would lie in a combination of the accuracy of your tooling and in your ability to use it. The sources for error that I see are:
X-Y Coordinates:
How accurate are the X and Y DRO scales on your mill? Any differences will put the holes on one axis at different distances than those on the other.
How square are the X and Y axis of your mill to each other. Any error will produce an angular error in the four right angle rotations: two will be greater than 90 degrees and the other two will be less.
Will you lock the X and Y axes down for each hole location? Will your moves from one hole to the next properly account for any backlash in the screws. I know you have DRO scales, but any looseness in the gibs can bring the table to different locations if a hole is approached from different locations. Don't believe me, just try it.
How accurately can you drill the holes using those scales? What technique will you use to locate them? For best accuracy with a mill I would mark, but not punch the locations. Then use a short, spotting drill to establish the locations. Then a short, stout drill at a smaller diameter to drill a starter hole. Then a drill that is a few thousandths small for the final drilling. And finally a reamer for finishing the hole. Yes, that is a lot of drilling. All these drills should be very sharp and as accurately ground as possible to avoid any deflections or other errors.
Rotary Table:
You don't say what size your table is. A 10 or 12 inch one would provide a lot more accuracy than a 4 or 6 inch one. Most rotary tables have Vernier scales that allow setting it down to tens of seconds with a basic accuracy also in that range. That should provide enough accuracy for your purposes. But you say yours is old so there could be wear.
Any slop in the center pivot of the RT would provide inaccuracy in both the X and Y location of the holes.
How accurately can you locate the table under your quill's axis?
How accurately can you locate the two plates on the RT?
To the above sequence of drilling and reaming, will your lock the RT down after each 90 degree move? If it is not locked, drilling forces can move the table within the backlash available in the worm gear and that can be a significant amount.
Will you lock the X - Y mill movements for each hole? Again, they should be approached from the same direction in spite of the presence of the DRO.
Personally, I would use my 10" RT as I do not trust the angle between the X and Y axis of my mill to be that accurate.
Note: I do not trust DRO scales to necessarily locate holes to their stated accuracy. This is not a mistrust of that stated accuracy but rather due to how they are used/mounted. Of necessity, DRO scales are mounted at a point that is offset from the quill's axis. So, any looseness in the gibs can allow the table to sit at an angle to the actual ways and this angle can be different with different sequences of movements prior to reaching a reading on the DRO scale. Errors of one or more thousandths can easily be produced this way. In a way, the scales on the handwheels which read the position from the rotation of the X and Y screws which intersect at or near the quill axis, are less susceptible to this kind of error. But I said "less susceptible", not immune.
For the best location accuracy using either handwheel scales or DRO scales, the gibs should be as tight as possible and that location should be approached with a sequence of moves that ensures that any cocking of the table inside the looseness of those gibs is always the same. That movement should be at least one full rotation of the handwheel, preferably more and always with the rotation in the same direction. Even a small backwards motion can destroy this accuracy.