Census records that are hard to find come in three categories:
(1) paper records that were never microfilmed;
(2) microfilm records that were never transcribed; and
(3) transcribed records that are not in the place expected.
Category (3) includes many cases where two distinct paper files have been transcribed under the same townland and DED headings in the database.
For example, in County Mayo, the paper records labelled "118/14" and "118/19/File 1 of 2" are both labelled "Sonnagh (Sonnagh, Mayo)" on the website.
As a result of this, there are duplicate database records for the first 28 houses athttp://www.census.nationalarchives.ie/pages/1911/Mayo/Sonnagh/Sonnagh/
Hence, for example, the "occupants" links for the five surnames under house 1 point to two different database records, one for the Kellys of Sonnagh (farmers), the other for Burke/Garahan/McCormach/Phillips of Main Street (drapers).
118/14 comprises the microfilm images fromhttp://www.census.nationalarchives.ie/reels/nai003025303/
The transcriptions from these images are correctly labelled "Sonnagh (Sonnagh, Mayo)" on the website.
118/19/File 1 of 2 comprises the microfilm images fromhttp://www.census.nationalarchives.ie/reels/nai003025647/
The transcriptions from these images are also labelled "Sonnagh (Sonnagh, Mayo)" on the website, but should be labelled something like "Main Street, Charlestown urban (Sonnagh, Mayo)".
This confusion arises most often in cases where a townland is divided into urban and rural files, which might be labelled, e.g., "Doonbeg" and "Doonbeg town", but are both labelled "Doonbeg".
I have reported numerous such glitches to the National Archives as I have come across them, but none of them appears to have been fixed. I could add more to this post if there is any interest.