The loss rate for manned space launches over the past 20 years has been 0.79%. What you're pointing out (only once every few decades) is a factor of the small number of manned launches per year. Actual loss rate is very high. 1/130 chance of loss.
Most (all?) of that rate provided by the shuttle program and by early vehicles from the 60s and 70s.
Mercury + Gemini + Apollo had a 0% death rate in-flight. Not that they didn't come very close.
The shuttle had numerous design flaws. The use of solid fuel. A booster that was assembled in seven different sections, fastened, and sealed with rubber o-rings. Go-fever. Foam insulation that was known to fall off and strike the orbiter during launch.
Not that space flight isn't dangerous. My point is we can do a lot better with proper designs and attitudes. SpaceX also gets to examine and improve their stages in a way that has never been possible now that they come back in one piece.
That you will always have at least single digit failures. Sometimes things have to go wrong before we can correct them. We got where we are today with airplanes not because the first million flights didn't have a single accident, but because we paid careful attention to every failure and did our best to keep it from ever happening again.
Take a look at this report on Boeing crash history: http://www.boeing.com/resources/boeingdotcom/company/about_b...
In particular, take a look at page 17 (slide 16) which shows the crash rate over time. We have gone from about 40 hull losses per million departures in the 50's to about 0.6 to 1 (depending on how you count) per million departures today.
