Hi Gerd and Henning,
If accuracy of srtm is +- 10m and the difference between with/without interpolation is +-1m, then it's definitely not worth spending any effort on interpolation.
This is not like interpolations error is placed within source data error. This error is added to data error. I believe it can be much bigger, especially in case of 3" HGT. This could be tested. Apart of height error, there is a problem of DEM shape. I mean the exact places where is a peak of a mountain or heights points in a profile of a route. With simple interpolation, there probably would be much bigger positional error. I have read reports, that increasing DEM resolution above HGT resolution increases DEM quality. But this was opinion about bilinear interpolation, maybe there is no need for such a settings with better interpolation? -- Best regards, Andrzej