They are planning to use SRTM data as the basis for construction purposes? That data is really coarse(1 value for a 25m x 25m area) and since all the elevation is integer instead of float you already have a fairly significant margin of error in the data(the actual value for a cell could have been 1249.49, but the sensor rounded to nearest integer), just based on how it was collected and stored(not to mention the SRTM in Eurasia has a mean height error of +/- .7m).
I don't think the focal statistics would effect the data to such a degree that it would cause significant changes in whatever planning they are doing. I'm guessing they are needed the contours for some kind of cut/fill analysis, what program are they planning to do the planning in? Many CAD-type programs can handle DEMs and generate surface contours from them all on their own. Actually usually if I'm making contours from a DEM I prefer to created them in AutoCAD/Civil3D as opposed to ArcGIS(for small scale areas/coarser data, if I'm working over a large scale or very fine data I use Arc, since CAD doesn't handle large surfaces as well as ArcGIS).
Dan_Patterson, curtvprice, got any additional comments? I was also thinking based on how coarse the original data is, would 5m contours even be "accurate" based on the large cell size?