Finally, I woke up in the middle of the night (yes, I dreamed d-scan last night) and realized that the "weird" short range phenomenon may not be weird but just a math problem that has to do with the scan cone geometry relative to the "distance" between the object icon and the warp in spot. I will see if I can do the math and get it to you.
You can do the math, but there is nothing magical going on, it's pure geometry as you suggest. The cause is that the planet's grid and bracket are at separate locations in space, sometimes by tens of thousands of km. From a trigonometric standpoint, this difference is ALWAYS negligible at distances greater than 0.5 AU or whatever, which is where you are when you are scanning from another celestial. This is the normal situation when you're scanning FROM another planet or its associated junk (moons, poses, belts, stations, non-standalone gates). Below those distances (those that are more normal when scanning WITHIN a planet's associated junk (planet, moon, belts, stations, non-standalone gates), it CAN (but not must, as it depends on the arrangement of things as well) start to make a difference. Stacmon 5 happens to be a great example, because from either the Impro Station or the Customs office (can't remember which), the planet 5 grid is actually BEHIND you as you're aligned to the P5 bracket. That is, things at zero to the planet only show up on 360 degree scan, not 180.
I wouldn't stress it too much, it's kind of a rare thing to be bitten by. Since scanning is an information theoretic game, it's useful to know all the rules you're playing by as you're binning locations into "target is at" vs "target is not at." This is just one of those extra conditionals you need to keep in mind, depending on how you scan. The bottom line is you will NEVER* be bitten by this phenomenon if you narrow your target to a distant cluster (planet plus its associated junk), then warp to the PLANET to continue your scan. If you feel that planets are a lower probability to catch targets at and choose to warp to another of the fifteen junk objects orbiting that planet (rolling the dice to get lucky at whatever you think the odds are) while you continue your scanning, you should be aware of this phenomenon.
Keep in mind that you can always "warp to" and cancel after you're mostly aligned to get the direction of the planet's grid from your location if you've been stumped and suspect this phenomenon is to blame. I believe GG mentioned this during the seminar, but using the align command on a planet aligns you to its bracket, not its grid. Therefore, a warp-to and cancel is required, aligning will point you towards where you want. This is, of course, just to show you where to point your camera due to where your ship is now pointed, I don't want to confuse things and imply that having your ship pointed somewhere somehow alters anything to do with the d-scan.
* I'm sure there are exceptions to every rule, and I certainly haven't visited all the celestials in New Eden to confirm this. There may be some planets that are so close (most of the orbits seem kind of uniformly distributed in log-space, so I'm thinking of scanning between the sun and planets 1 or two/vice versa) that this actually can make a difference.