Interesting post drambuster.
One thing is that the NASA data is crippled in accuracy outside the USA. I recall reading something on the website more or less saying this... 300ft accuracy outside, 100ft accuracy inside or something like that.
There were many complaints from many countries about NASA publishing data enabling the countries to be attacked by their usual enemies (in the 3rd world this fear is pretty standard). The visual imagery (initially made available widely on NASA World Wind) shows loads of "top secret" military installations in the 3rd world etc etc. We will never know how much data has been crippled. But the public imagery is definitely for the most part years old. Of course this time delay doesn't explain missing terrain!
However, the more basic thing is that only a fool would rely on these terrain depictions to fly in IMC close to terrain. [E]GPWS is a last resort lifesaver only - coming into action if the normal situational awareness picture has come apart due to some major pilot errors - and nobody I know of pretends otherwise.
If SV is to be used for actually flying an approach, some serious database supplementary work will need doing for the immediate area. An ILS or some other precision approach with a DH of 200ft takes you within a mere 200ft of the ground
Even the Shoreham 20 NDB/DME approach, when plotted onto a 1:25k O/S chart, takes you within a few hundred feet of the terrain a few miles out (this is obvious when you fly the GPS track and the vertical profile precisely, autopilot-coupled, and look out of the window). Only the "mandatory DME" stepdown stops you flying into the hill if you suffer a fairly normal NDB lateral error.
I have tested my Garmin 496 based "EGPWS" setup and found the warnings to be fairly accurate in all cases. But I accept there are major errors. American pilots have already reported totally spurious warnings.