First I have to say, it's not easy to orient myself in this forum, despite I am quite familiar with forums. In this I feel lost. So I put my text simply here, feel free to move it to a 3d-Analyst subtheme (which I didn't find).
I am on ArcMap 10.4 Standard, with 3d-Analyst.
Arcmap itself has the "Identify"-tool, and when you click somewhere in the map, where is a raster dataset with elevation information (32-Bit floating point, for instance) present, you get back height information from that layer, which is the cell value of the raster pixel located there. So far, so good.
With 3d-Analyst, you have the two functions "Add Surface Information" and "Interpolate shape", where one can transfer elevation data from a surface like the raster dataset I mentioned, to a shapefile, either as information in the data table ("Add Surface Information") or directly as a new z-enabled feature class.
Recently I detected some differences between results from those 3d-Analyst-operations and results from the tool. This means, I have a point shapefile where surface information was added, but when I click on some of the points with the identify tool, the elevation from the source raster is not the same as the points got from the 3d-analyst operation, however near I zoom in. Of course the underlying raster dataset is always the same.
For the kind of surface I use, 3d-Analyst forces me to use bilinear Interpolation, although I rather want to use linear, or the method Arcmap itself uses to "Identify", which would be simply the value of the underlying cell. In the cases I have, I do not need any interpolation, because manually going through the results shows differences between what's there in the dataset, and how it was calculated.
So I kindly ask the developers (and the community if you have a way to come around this) to either enhance the "identify" tool, or to provide a non-interpolated mode for the 3d-Analyst functions, since restricting the bilinear interpolation via reducing the "sampling distance" to a very small fraction of the raster cell size does not work as expected.