Been sitting on this one for a while because I wasn't sure it was worth posting, but yeah - I've noticed the same pattern.
There's an old disused line running through part of the New Forest not far from me, Victorian-era construction, and the local accounts I've collected over about six years consistently cluster within maybe 400 metres of the old trackbed. Three separate witnesses, none of whom know each other.
My working theory is that it's not the tracks themselves but what the construction disturbed. Railway navvies in the 1800s cut through some seriously old terrain - burial mounds, ley intersections, ancient woodland that hadn't been touched in centuries. Whatever these things are, I think they're responding to a disruption that happened 150 years ago and never really settled.
The American reports show the same geographic clustering and I don't think that's coincidence. Has anyone actually mapped their local sightings against historical OS survey maps to see if theres a pattern with the original construction routes?