I just found the exact same thing last week. I got across an 18in wide plate, and when I turned the fixture 90 degrees I noticed the holes didn’t line up. The row to row spacing was consistently short by about 0.003-0.004, which left row 7 short by ~0.025 relative to row 1. The plate is 18 x 52 but the long edge is dead straight compared to the short sides, so I used that as the starting edge, and was going to finish the plate down the length with the template at 90 to the original pattern, but this kinda screwed up that plan. Before going any further, I decided I wanted to figure out exactly what was going on so I got out the inspection tools.
I found that the 2nd row on my template is a bit short by 0.001, but that alone isn’t going to cause this issue. From what I can tell, the major error is coming from the tacking bolts since they measure 0.746" diameter and don’t have a feature to keep them centered in the template bushings. The lock balls should help center the pin in the table hole a bit, assuming any burr left from the cutter is consistent all the way around the hole, but that could be another source of error.
I’m guessing the grooves on the tacking bolts are for o-rings to keep them from falling out of fixture blocks when being moved around, so I put a 3/4” x 5/8” o-ring in the upper groove that sits in the template, and that helped the alignment quite a bit. Unfortunately I just have “generic” unspecified hardness o-rings on hand, so I could still get some movement in the template before tightening the bolts down, but a higher durometer o-ring would probably help stiffen things up a lot more.
To go one step further, I bought some 3/4" dia cold rolled bar (0.750 +0/-0.002) and used it to make additional alignment pins. They are almost “too tight” since my holes are coming out right around 0.7505 and if there’s ANY amount of debris left in the hole or template, they bind up or just won’t go in at all. So staying consistent with blowing out the holes out was key. This way the tacking bolts were primarily just for clamping the fixture plate down.
To continue down the plate, I used the middle row of the first pass as my “zero” point, instead of the edge of the plate, and extended that row down the length like this:
The cold rolled pins and careful measurement from the good edge helped keep the template located while extending that row, and then I was able to use that row to make sure the spacing stayed accurate as I drilled the rest of the plate.
For the hell of it, I also tossed a couple tooth blocks into the mix so I was referencing more than just the previous row of holes, and the hole pattern across the rest of the table came out dead on perfect! I used the big corner jig to test the pattern, and even without chamfering the holes yet, it drops straight in. 
I put things back together to take this picture for reference, but I have more plates to drill so I’ll take pictures of the process I came up with when I start the next one. I also used a clamp on the far end of the template, but just didn’t put it on for this picture.
I still have to go back and fix the original part of the plate by reaming the holes with the template at 90 degrees, so I’m going to use the middle row as “zero” rather than the starting edge. This will help spread the error out across that section and keep me from having holes that are “ovaled” by almost 0.030.
