Those are additionally generated nodes, their numbers have to be different than the numbers of nodes of the mesh.
Currently, reference points are only displayed and can’t be selected for querying. To apply loads or BCs you also have to select them from a list, not by clicking directly on them.
Coordinates of the ROT NODE can be arbitrary and in PrePoMax they are the same as those of the REF NODE so these nodes coincide.
When you apply a boundary condition or load to a reference point and select rotational degrees of freedom then that BC or load is internally applied to a ROT NODE (but the numbers of degrees of freedom are 1-3 like for translations).
You can have gaps in the numbering of the nodes. If I remember correctly they are not advised as they increase the memory consumption of the matrix assembly (maybe), but they do not affect the memory consumption for the system solving. Since system solving memory demand is much larger, the assembly should not affect the overall memory consumption.
If you have 2 missing nodes, this is probably due to an additional reference point which you are not using since each reference point needs two unique node numbers.
The reference points cannot be queried. Their node ids change every time you generate a new mesh so it could be misleading to show the node ids for the first mesh and then after remesh the node numbers would change.
The CalculiX documentation states that the position/location of the ROT NODE is irrelevant so it makes no sense to show it graphically as a separate node:
To apply the load or BCs to reference points you use a drop down list: