I was surprised by this as well, mainly because of the mismatch between commonly available bathymetric data and the reported depth of 1400 m at the site of Sandy Island. However, I think the reports are correct, there really is no Sandy Island at that location (even though it's been mapped as early as 1908, http://blog.auckland...hat-isnt-there/
The research mission was undertaken by oceanographers using a research vessel that is well equipped for bathymetry measurements (http://www.marine.cs...edocs/index.htm
, voyage number SS2012_v06). It's very unlikely that they would get their location and the depth soundings wrong.
Also, if you just take a look at the Landsat images of the area, you'll clearly see that there's no island. The Landsat images should be considered most reliable because unlike the Blue Marble images they aren't composed from many different images using a Land/Sea mask. You can browse the images at http://landsatlook.usgs.gov/
: Navigate to the area, zoom in on the island (which is present on the orientation map), and click "Select Scenes". You can compare images taken on different occasions, and they all show empty ocean.
Of the sources you quote, two aren't independent: World Vector Shoreline and World Databank II are both derived from other maps and charts, and it's very likely that they used the same source for the region in question.
The NASA Blue Marble imagery shows something that could be taken to look a bit like an island (http://www.blue-marb...5...ay=0&base=0
), but I think it's just an artefact stemming from the fact that the land/sea mask to generate the composite image also includes Sandy Island.
The bathymetry in the Smith & Sandwell global grid at the location in question comes from satellite altimetry, the most indirect and imprecise form of depth measurement. There is no multibeam ship track at the location. So I'd say that the bathymetry in this area and in this dataset is just wrong.