I don't know what we can reasonably expect people to assume when it comes to teleports, since there is no real-life experiences other than maybe other games, and I'm sure it's not very consistent from game to game either.  I haven't tested it, but I'm guessing Lemmings 2's teleports are probably skill-canceling and outputs to same location.  On the other hand, although I haven't played those games at all, I have a feeling those portals in the "Portals" series of games probably behave much more like what we have here, where a lot of state get transferred directly to the output.

I think it's more important to understand if the current behavior are actually creating situations where the level author and level players can easily shoot themselves in the foot?  That is, the level author isn't intentionally making use of, say, the output offset behavior, but the behavior introduces unwanted precision for the level players in a level, and at the same time the level author can't easily tweak the level to remove the precision?  This would make for more convincing arguments than corner cases, otherwise it feels a little like hidden traps, where we've simply been "handling" things by telling people they are lame.

I also think that for discoverability, it would be easy enough to create an introductory level for teleports that directly require and illustrate these two behaviors, by making it obvious in the level that the player have to, say, use one single builder who gets teleported mid-build.  It would at least directly show the skill-preserving behavior, and the more observant may also see the offset-preserving behavior as well.

It may also be interesting to think about consistency, even though there is no direct requirement for unrelated things to share similar behaviors.  For example, blockers' effects on other lemmings are at least skill-preserving in most if not all cases (not sure a clear analogy can be made there for the output offset aspect).