Like Joan_Sparky said, you probably have to accept this as a bit of quirkiness that helps make KiCAD so "unique", "versatile", "charming", "weird", "thought provoking", etc. (Think, for example, of all the different adjectives my wife uses to describe me, depending on what day you ask.)
There probably IS a reasonable explanation in the realm of data and "virtual objects". It appears that in the KiCAD universe, "holes" don't exist as objects in their own right. They are always associated with a copper "pad". In the grand scheme of things this may be either a wise decision or foolish, but for now we must accept that it pleased the Creator to define the universe in this way.
We must also accept that a "pad" must be associated with at least one copper layer. Yeah, this contradicts what we encounter in the "Pad Properties" editing box. (Can somebody give an example where the "None" option is a correct selection in the "Copper Layers" section of the "Pad Properties" dialog? Perhaps it is time for a bug report.)
To create a simple, basic, hole in the KiCAD universe we must define both the hole and its associated pad as having the same size. In the physical world where we live and breathe this is a degenerate case - the hole completely swallows the pad, leaving nothing. (To begin with, since a hole is a place where there is nothing, are we arguing about nothing?) But in the KiCAD universe, both the hole and the pad are virtual objects, they can exist simultaneously, and neither can annihilate the other. Other EDA programs I've used would squawk this degenerate case, citing something like "inadequate annular ring", but KiCAD is more tolerant and less legalistic in that regard.
Even though the KiCAD hole has reduced its associated pad to nothing, that nothing must be placed on a copper layer. And in KiCAD, a hole goes through all the layers so we must specify "All Copper Layers" for its associated pad as well. I know a few good machinists who can make half a hole rather than a whole hole but they obviously aren't allowed into the KiCAD universe - only complete holes, leaving nothing on any copper layer. Now, since we must specify "ALL Copper Layers" for the pad attributes associated with a simple, basic, hole . . . shouldn't we also be required to include ALL of the "Technical Layers" in the pad definition?
Is there a less convoluted justification for the definition and behavior of simple holes? Perhaps it's time to rethink the way holes are defined and used. I see that selecting "NPTH, Mechanical" automatically disables the "Net Name" and "Pad Number" input boxes - could selecting "NPTH" also force the selection of other options? In most cases there won't be any copper ring around a NPTH, but occasionally you might want a copper ring around the hole as either a chassis grounding connection, or for mechanical reasons. (The slightly elastic copper absorbs vibration and thermal expansion under a fastener, where brittle glass-fiber board material abrades away.)