It's important to notice the difference between 'resolution' and 'posting'. We tried to get the most resolution from the hardware and it ended up being about 30 m. For creating the data sets, we settled on postings of 1 arc-sec as that was near 30 m near the equator. Thus we didn't have to project the data. For areas outside the US, we averaged to 3 arc-sec. There is more information in the publications listed at http://www2.jpl.nasa.gov/srtm/srtmBibliography.html

SRTM and other shorter wavelength radar systems don't penetrate heavy vegetation. The elevations SRTM shows for those areas are near the top of the canopy. There have been a few articles published quantifying this effect.