The operator at Station 1 performs a measurement on a particle there that changes the spin of that particle, as well as a measurement on a desired quantum bit that we want to transmit to Station 2. This process produces 2 classical bits of information. This process destroys the correlation between the particle pair, and affects the particle at Station 2 but in a randomized way that is useless without the operator at Station 2 possessing the two classical bits (thus it is not faster-than-light communication).
Once the laser takes the two classical bits to Station 2, at the speed of light, the operator at Station 2 can use them to decide what measurement to perform on the paired particle. Once the correct measurement is performed, the quantum bit that we measured at Station 1 appears.
The security benefit is that by completing one risky journey from Station 1 to Station 2, we can allow many future quick, secure journeys.
The propagation happens at the speed of light because the classical bits are required to understand what has been teleported. Performing the measurements is much slower than the speed of light. The correlation can only be stored for very short periods of time, currently.
It also bothers me that the article specifically says the "information" is teleported, as it seems like the author is going out of his/her way to imply a violation of the No Communication Theorem (http://en.wikipedia.org/wiki/No-communication_theorem), which is not in any way violated by the experiment.