To the best of my knowledge, there are no known terrestrial populations of A.mexicanum; and where two or more species of the complex come into contact, they differ in morphology, ecology, and genetics, and for the most part do not interbreed. A.tigrinum and A.californiense in particular have not shared a gene pool with other members of the complex in millions of years.
Neoteny in these species may be obligate in some, nearly obligate in others, and facultative or absent in yet others. Neoteny in itself is not the reason for recognizing these species, although it is a factor for some. There are multiple lines of evidence
To assume In any case though, whether they might merge in future is purely hypothetical, and we must deal with the evidence of here and now. To do otherwise might require us to begin naming siblings to separate genera because their hypothetical descendants are so assigned. It's also a false assumption that a surrounding terrestrial population exists. Since the non-neotenic animals reproduce in the same places and ways that the neotenic ones do, if there ARE such surrounding populations, they either do not interbreed for some reason and are thus separate gene pools and separate species; or they DO interbreed already, in which case there's no point in even discussing them, since they are already a single interbreeding population. The essence of ALL species concepts is that a species shares traits by virtue of a shared reproductive pool. If the unique set of traits of one population is not found more or less throughout another population, despite some degree of interbreeding at their contact zones, then they are not sharing a gene pool, reproductive pool, or ancestry, and are not the same species. This is true of all currently named Ambystoma I know of, and in fact also true of several populations currently assigned to A.macrodactylum, A.mavortium, A.californiense, and perhaps A.tigrinum, A.maculatum, and a couple of others as well [the latter few requiring much more data in comparison to the first].
I do not see what flaw you refer to in current species definitions, but I definitely see flaws in past definitions: whereas definitions used in past were often in disagreement and used entirely different measures, current methodologies have a very strong tendency to identify as species, organisms which could retrospectively also be defined as species by MULTIPLE past methods. They all share in common the fact that the traits we observe and measure are mostly a product of speciation, not a cause; and they are a product of common reproductive pools. Logically then, if we can identify a common reproductive pool more directly [ie, via genetic measures], then other indirect measures can be considered helpful but unnecessary. The result is, especially among salamanders, that we can have separate species which appear by any other measure to be identical. In reality, we can often detect additional 'invisible' traits, such as ecological niche, pheromones, diet, etc, if we look carefully enough. It doesn't matter really whether you're comparing Ambystoma lermaense to A.taylori; or A.laterale to Hynobius tokyoensis: in both cases they simply do not interbreed, do not share recent ancestry, and despite outward similarities, should not be considered conspecific.
I know that concepts and understandings were different 30 years ago when I first started studying these subjects [although some of the techniques now in common use were in fact being applied 50 or more years ago], but since then I have continued to follow the research on an essentially daily basis, on a wide variety of taxa. This has given me a good view on how and why changes have taken place, and also allowed me to see some flaws and even very common errors made by those who know the subjects best. Generally, the conflicts arise because the methodologies and logic involved are not fully understood, and because people get very stuck in their viewpoints, regardless of the evidence. I'm not a big fan of indistinguishable species, but I do recognize their reality.