I read that as the raw SRTM data being roughly a 60m x 90m grid but interpolated into something finer and then smoothed after considering a much wider area (300m x 450m). I can't see why it would use the term 'subdivided' if the data was at a ~1 arcsecond resolution anyway. Have I misread this, are the SRTM raw data are at a ~25m resolution? Is there more up to date SRTM data than the data this page is talking about?
No - SRTM 3 covering the UK at around 90m resolution is the most recent as far as I am aware. The ~25m (1 arc-second at our latitude) figure I mentioned comes from the fact that at the pre-processing stage, I think JPL were using this resolution to then derive the 3 arc-second data that have been made public. They released the 1-arc second data for the US, but for the rest of the world that is not publicly available. I mentioned this because the significant resampling and filtering that has gone on is what has ended up removing the human-made features from the DEM. For some of the details, see
SRTM_Topo.
OS OpenData offers a "gridded digital terrain model with a 50 metre post spacing" which is derived from the contours on 1:50,000 scale mapping. It sounds interesting but I've no experience of it or whether it could give results closer to contour counts based on 1:50000 OS maps.
It should certainly give results closer to 1:50k contour counting than just about any other DEM source. When I have time, I will look at exactly how similar or different they are. This is a subject close to my heart as part of my
PhD thesis involved looking at the accuracy of the 1:50k DEM by comparison with contours and examination of erroneous artifacts. The work is nearly 20 years old now, but it uses the same DEM that is currently available via OpenData.