Here's a coupe of suggestions to consider:
1. Try sniffing late at night as a number of the 40KHz emitting noise distractions may be dormant. That may decrease some false returns.
2. If you have the ability to attach a portable oscilloscope to the audio output port (I use scopes in my vehicle powered by a 12VDC to 120VAC sine wave inverter), you will be able to visually discriminate between the various ultrasonic time-domain signatures. Despite other interfering signatures, you should be able to see the 8.3mS arc-fault signature on the scope. Using one's ear to discriminate between different signatures coming from the ultrasonic receiver is not very accurate.
I have some other suggestions but those would require modifications to the MFJ-5008. If interested, email me per my QRZ email information.
Good hunting..