The subject title covers the question already.
I was under the impression that a leak test checks if a computer can send out information to the net - ie outbound traffic not generated by the user.
If a leak test fails - ie reports that traffic was sent from a computer that’s not even connected to the net, there is something strange going on…or do I not understand what a leak test is supposed to determine?
It is possible to create a leak test that would leak frrom one PC on a LAN to another PC on the same LAN without using the internet.
In the case of a personal firewall installed on a PC, the firewall is supposed to stop data being transmitted autonomously. Whether it leaks outwards to the internet or the internal network, it’s still a leak if the data leaves the PC without user or firewall permission.
To test a LAN based leak test, you would need to modify the destination of the leaked data to a local LAN address and have some sort of an agent on the receiving PC to accept the incoming data.
Cheers,
Ewen 
PC Flanks Leaktest succeeds when you’re off-line because it considers itself successful if it manages to send a message to MSIE (which CFP allows it to) rather than MSIE returning the message to their web site (which CFP doesn’t allow if the user says no).