Let me remind you that the "rocketing" of an aerosol can is very much a secondary issue. What is important is the momentum-balancing travel of the liquid, providing that liquid is combustible. Obviously, is the liquid is salt water, neither the trajectory of the can or the water is of much import.

In the K-Mart fire they were dealing with a carburetor cleaner. Let's look at the formulation of a contemporary carburetor cleaner aerosol:

METHYL ALCOHOL* 1 - 10 200 PPM FLAMMABLE
XYLENE** 15 -25 100 PPM FLAMMABLE
TOLUENE** 15 - 25 100 PPM FLAMMABLE
LIQUIFIED PETROLEUM GAS 25 - 30 1000 PPM FLAMMABLE
DIACETONE ALCOHOL 1 - 10 50 PPM FLAMMABLE
ISOPAR M 20 - 30 NE FLAMMABLE
METHYL ETHYL KETONE 1 - 10 200 PPM FLAMMABLE

The rupture pressure of the can is about 200 PSI. This pressure can be reached in a car parked in the sun so it can certainly be reached in a fire scene without the actual flames being anywhere near the can. Once the can pops, the LPG vapors will provide the path for the flames to follow to the displaced liquid. It really doesn't matter much how much accelerant lands on a distant target (such as a blanket). A little accelerant can go a long way in more than one sense.

You suggest that we wait until someone conducts "scientific" tests to show that a spray can of this type can create a second apparent origin. I suggest that fire investigators remember that it is their job to eliminate reasonably possible alternater causes. Thus the burden properly lies with the investigator to conduct any research necessary to positively eliminate spray can effects before declaring multiple origins.

The problem, of course, is that if an investigator has already summarily dismissed aerosol cans in some previous investigation, it may be difficult to persuade him that he may have been to hasty.