I don't see why this is inherently more dangerous than, say, flying a powered parachute. It's really more in the same class as a helicopter except that engine failure risk is mitigated by the emergency parachute system. Running into stuff while flying low is probably the main mode of death.
Mainly due to the fact that a 'powered parachute' (e.g. an ultralite) undergoes essentially no mass modal volume delta in airflow when transitioning between powered and unpowered flight.
The controls, and fundamental aeronautical responses to flight inputs, remain the same in both flight regimes and, for the most part, there's a gradual and manageable amount of pilot feedback during the transition.
In fact, this lack of "gradual feedback" was the main reason that power steering took so long to be safety-approved in automobiles. It's also why, in most countries, pure 'drive-by-wire' controls are only allowed in experimental and prototype road vehicles. Despite the obvious ergonomic and cockpit layout virtues of joysticks, steering wheels and worm gears can still work without active power-assists.
The same can be said for a gliding light airplane or helicopter under auto-rotation.
Alas, the same CAN NOT be said for the Martin jetpack when deploying its ballistic parachute. Btw, aside from cost and liability issues, this is one of the reasons emergency ballistic parachutes haven't caught on in small single-engine aircraft.
I wonder how much altitude you could expect to lose from the moment the engine quits until the moment the parachute is fully deployed and slowing your descent? Surely it would be over a hundred feet - which means that any engine failure in that altitude range would result in a painful splat on the pavement. http://en.wikipedia.org/wiki/Height-velocity_diagram
This could deploy more quickly since it's actively assisted by the explosion. This shoots up, rather than floats up (admittedly "float" in this case is a pretty high wind velocity! ;)
A skydiver is transitioning (abruptly) from unpowered free-fall into unpowered, albeit greatly slowed, descent. The jetpack user is transitioning from powered, controlled horizontal flight (with its characteristic 6 degrees of freedom) into unpowered descent.
A back-up chute for the skydiver has essentially, but not quite, the same flight characteristics and initial deployment conditions as a main.
Parachutes in a stable flight configuration are incredibly safe and reliable, but it's GETTING that 6-axis dynamic flight-body INTO that stable flight configuration that's the real challenge.
It's relatively easy to design a system for even extremely large, static, dead-weights (such as cargo or even battle tanks) to be safely launched, arrested, and landed via parachute.
It's an entirely different problem when your safe chute deployment must take into account the wide variance in airspeed, pitch, yaw, altitude, and environmental influences which powered free flight allows. Imagine a worse-case scenario where all power was lost just after a major inadvertent yaw maneuver (it appears to be designed to minimize yawing). That's a sticky-wicket to solve repeatedly and reliably.
In the case of the jetpack the parachute is launched upward with rockets. This takes extra time: the time it would take for the operator or system to detect engine-out and initiate parachute launch, then the time it would take for a rocket(s) to launch the parachute up to where it could be, then the time for the uprush of air to catch the canopy.
Here's a link to an article about the ballistic parachute fitted to Cirrus SR20/22 aircraft. http://www.cirruspilots.org/content/CAPSHistory.aspx There's one mention of a low-altitude parachute deployment: "one activation occurred at too low an altitude to fully inflate the canopy (witnesses report 200 feet above ground), and another activation where the rocket took an unusual trajectory resulting in a failure to open the canopy."
I think the problem is that the airframe provides no tendency towards being aerodynamically stable, the flight stability relies on good pilot input, thus tending to be less robust and forgiving in situations involving pilot error or mechanical problems.
