A couple points:
LiDAR does NOT survey every single spot in a location. It is a series of points, and if there's a gap over the actual highpoint, it could result in an artificially low peak value. It could be significant if the summit is particularly pointed or has big boulders. The only thing I can confirm is that 3995'
is the highest point in that LiDAR set. I haven't taken a close look to see how far apart the gaps are.
Secondly, Roy S. has pointed out elsewhere that the old disc was not quite on the true summit. Almost, but not quite. As far as the datasheet for that benchmark is concerned, that was last reported as destroyed. So I would hold off any opinion until it's confirmed the new one is absolutely on the highest point, and the report on the survey methods comes out.
I'm intuitively surprised to hear GPS can be more accurate than LiDAR except perhaps in the thickest of woods, so I'm guessing the LiDAR absolute (vs. relative) accuracy is off a lot more than it should be. That, however, can be corrected with control point surveys (at least that's what I took away from some recent reading).
Worst case (if one has a distaste for accuracy

) is it'll still be on the NEHH list, we'll be even with the DAKs in numbers***, and gridiots will have an easier time of it.
*** A rudimentary analysis I did besides Tecumseh shows Lincoln and S. Hancock should go away due to lack of a 200' col, and Guyot should get added. Thus, 46 peaks.