There are several reasons why this can occur:
1.) A particular “benchmark” was never part of the National Spatial Reference System (NSRS), to begin with. Perhaps another federal agency, even the Corps of Engineers, established the elevation of a benchmark but did not “bluebook” the observations and descriptions nor submit the data to the National Geodetic Survey for incorporation into the NSRS.
2.) A benchmark was originally published on the NGVD29, but was not included in the readjustment to the NAVD88 because it was not observed (re-leveled) in South Louisiana during the 1980s through the early 1990s.
3.) The published elevation has been deemed by the National Geodetic Survey to be unreliable and has been “pulled” from the published records in 2006.