What Jim is seeing, is the same thing I see, when going into the abyss. (Abyss, is a valley, covered with tall pines, and other obstructions).
Quite simply, it is where the obstructions, and the and satellite configuration are coinciding. Back in my L1 days, these sites could be gotten, with mission planning. In our modern times, with the Javad, another shot later in the day, when the Sats have moved, he could get it with LOWER RMS.
This equipment is STILL a statistical measurement system, and subject to all the things that interfere. So, if this were an important shot, then do it 3x.
I like to think in terms of a good camera, with adjustable exposure. When all things are very still, you could set the camera to a LONG exposure time. Well, setting the LS to a 3 minute exposure, and it "Gets it" but has higher RMS than desired, it simply means "This is the best I could get, during this particular 3 minute window". Shoot it 3x more, at differing times, and then use the AVERAGE function. (I am sure Jim is probably AHEAD of me on all this) but the AVERAGE function typically will ALSO weed out the higher RMS shot, and will generate an average shot, that is CLOSER to the ones with smaller RMS. I have been playing with this very thing.... Remember, I try to BREAK things like this, and find it's weak points.
I will also add, when a shot generates a "string cluster", where the 180 seconds of data shows a string of dots, with not alot of clustering, those are often poor shots. I had one of these string shots, recently, 180 seconds, and re shot it... the final coord was some 0.085' from the "String shot".
Some of what I am writing is my "Impressions" from a USER point of view. Somebody more technically oriented could probably say all this in a "more better way". (That's a southern way to say better!)
Carry on...
N