After making my best efforts to read/try everything I can on this forum and elsewhere, I unfortunately have to contribute to the wide literature of "my layers aren't aligning" questions.
I am working in SW Ontario & have some somewhat reliable reference data in NAD 1983 CSRS (Canadian Spatial Reference System). Then I have two shapefiles of points:
1). One that I created from a raw spreadsheet of lat-longs. The source reports (somewhat half-heartedly) that it should be in NAD 1983, so that's the geographic coordinate system I've defined for it.
2). One that was sent to me already as a shapefile, projected into NAD 1983 Great Lakes Basin Albers. I think it's quite possible this was not the data's "original" coordinate system, but I don't know. I've gotten the original raw data and can try to plot it myself. Ultimately I want everything to be in Great Lakes Basin Albers.
Between these 2 datasets, certain points are supposed to be redundancies. When I hone in on such points, they have the same lat-longs (in the fields that were used to plot them), but the point from #2 is always 13-23 meters NE of the point from #1. I say my reference data (streams) is "somewhat reliable" b/c, while I know it's projected right, the points aren't necessarily going to be right on top of the stream lines, though the points from #1 often are, while #2 is offset. Thus, I've assumed my streams and #1 are okay.
This kind of looks like a central N. America NAD83 / NAD27 offset to me, but when I reproject #2 to a geographic coordinate system of NAD27, it only gets further away to the northeast. Throw in the datum transformations needed when the CSRS version of NAD comes into play, and I'm quite unsure how to diagnose my problem in a systematic way or fix it, i.e. to get these points with identical lat-longs to actually be on top of each other, preferably at the location of #1.
EDIT: The #2 points appear to have an attribute for "Geodetic Datum" specifying either NAD 1927, NAD 1983, or WGS84 (& very often blank). They're museum records covering roughly 1930-2010, so this kind of makes sense. Does this mean their lat/long attributes are all based on different reference points & the shapefile can't be fixed in a uniform way?