I haven't personally done a store locator, but...
Store locators are relatively crude, yet useful. It is unlikely that
exacting math will make one less crude in such a way as to make it
significantly more useful. They usually just display a list of locations
that *might* be convenient, and that's probably "good enough."
I'd guess that a majority or potential users wouldn't know their Z+4,
wouldn't type it in if they did, and would be unwilling to give
address/phone info, except when required for travel directions.
Even if one were to know the user's exact coordinates, the distance between
coordinates is crow-fly distance, not travel distance, and not travel time.
Rivers, mountains, congested areas, and so on make exact distance a crude
approximation of location convenience.
As to the recommendations for (U.S.) databases -- I think they're all
derived from the Census Bureau's Tiger database -- even the products from
the Post Office that provide latitude/longitude. The PO version probably has
good updates, but the Census Bureau releases theirs every couple of years,
presumably with PO updates, and it's free (and large). I don't know if any
of the commercial versions are improved in any way.
As to the concerns over the varying coordinates between web sites, they're
probably all correct (from a quick look at some of the differences posted,
they appeared to be within 1/4 of a mile). Other than what appears in the
Tiger files, there is no "official" coordinate for a zip code, because zips
aren't points. They aren't even geographic areas (like a city or county with
boundaries). They're just networks of delivery destinations.
The Census Bureau developed Zip Code Tracking Areas (ZCTA) as approximations
to the geographic areas covered by each zip. The listed coordinates are a
calculated crude geographic center for each of the resulting irregularly
Given all the approximations and that ZCTAs can be hundreds of square miles,
it's obvious that any derived location information is crude at best. Zip+4
would indeed be less crude, but still problematic and likely no more useful.
Probably less than $.02 worth.