I reread the original schmidt reference, and he IS talking about the Loltun Caves, which is what I originally thought. So we're not conflating two different issues - it is one and the same, and it has been debunked.
Incidentally, horse bones were also found in association with cultural remains at Loltun Cave in northern Yucatan. There, archaeologists identified a sequence of sixteen layers numbered from the surface downward and obtained a radiocarbon date of about 1800 BC from charcoal fragments found between layers VIII and VII.66 Significantly, forty-four fragments of horse remains were found in the layers VII, VI, V, and II—above all in association with pottery. But the earliest Maya ceramics in the region date no earlier than 900–400 BC.67 Archaeologist Peter Schmidt notes,
What clearly results is that the presence of the horse, Equus conversidens, alone is not sufficient evidence to declare a stratum totally Pleistocene given the long series of combinations of this species with later materials in the collections of Mercer, Hatt and others. Something went on here that is difficult to explain. [Difficult to explain, that is, in light of current theories about the extinction of the pre-Columbian horse.] If a late survival of the horse and other Pleistocene animals is postulated as an explanation of the situation, it would have to be extended almost to the beginnings of the ceramic era, which will not please the paleontologists.68
These ARE Mercer's horse bones that were found to be modern.
It would still be interesting to see the Schmidt source, to find out if he made a dating error that was later corrected, or if Sorenson misused yet another source.