M1 has a more complex data pattern that would be difficult for noise to produce.
Actually, the data pattern transmitted on the powerline is comprised of the data and its compliment, preceded by a 1110 start pattern. So M13 on the powerline is actually:
1110010101010101010101
A burst of 120KHz just after a zero crossing of 60Hz is decoded as a logic "1". If the noise is slightly different on positive and negative half cycles of 60Hz, and the noise amplitude is just right, it is possible for it to create the 010101... pattern. Then if there is some transient that interferes with the background noise in such a way that it creates the start pattern, it is possible for a X10 module to decode the noise as M13. I have witnessed several instances where my set of noisy CFLs used for testing were able to create a valid X10 command this way.
Another likely pattern is J Status Request: 1110101010101010101010
That only needs a little extra burst of noise added to the background to produce the start pattern, and I suspect that is what my CFLs were able to create.
Jeff
