@Trobec First, apology for the (very) late reply. In the last few weeks I wasn't master of my time, and on the forum the topic quickly sinks into the depths.
The data on MaPZS (are still) not based on Lidar. PZS does have all the data, including Lidar processing, but it hasn't made it to the maps yet because it's not that simple. It will, sooner or later. PlanGIS (internal version of MaPZS) already has Lidar heights and saddles as one of the layers. The difference between MaPZS and individual maps is that they come from different sources, so they differ from each other. Even "official heights" are in places decades old, measured with the technology available at the time, according to the standards then in force. And since in many cases it's hard-to-reach points (not necessarily technically difficult, but more in the sense that it's hard to drag a theodolite through the Alps, not to mention the motivation for such an act

) no one ever felt the need to check them.
Lidar heights often differ from OpenTopo and similar maps. Off the cuff, I'd say only about half of the heights are correct, and another 30% are in the +-1 m range, which let's say is about the same.
Or put differently, every fifth peak has noticeably wrong data on the maps. There are 4 types of errors:
- height doesn't match (mostly by a couple m, but there are cases with dozens of m difference)
- location doesn't match (official peaks often not at the highest point)
- peak exists in nature but not on any map (especially in mid-mountains, i.e. on Snežniška Planota, in Kočevski Rog... full of such cases)
- peak on the map but doesn't exist in reality (viewpoints, ends of ridges, geographical names of some area that actually isn't a peak etc.)
@lijaneja Vrh Nakla is a typical such example. It's a contour line with exactly 2000 m height and only 14 m above saddle and marked as peak on many maps. Lies right by the marked path on Rombon and ascent takes a minute. But somehow it ended up on all lists and since it sounds cooler to conquer e.g. 5 peaks in one trip instead of just one, it persists there.