Rather than redefining the parameters for an evacuation test, how about using real-life data from the last 10 years? There's been quite a few evacuations, following different kind of incidents and in wildly different settings. 'Real life' data will always give us a better picture of how passengers and crew will act than any simulation, regardless of how well thought out it may be, so why not use that data to see if there are any reasons for change? It should start with a couple of very simple questions: Was evacuation achieved within 90 seconds? If not, what caused the slow down?