This may be of interest, report from the Air Accidents Investigation Branch:
The drone took off from a road bridge over the railway, rapidly lost control and crashed. Investigations pointed to magnetic interference from the power cables above the railway but below the bridge, demonstrated afterwards to significantly affect an ordinary compass, as well as showing up on the flight logs.
Anyone want to add a compass to their standard site kit? 
Bizarre… I’ve flown over, under, near powerlines hundreds of times in the past.
There are also some videos here of members flying through the power lines 
That’s where your magnetic interference is most likely to come from. Either REBAR or steel structure.
Could be something to do with magnetic relationships between the cables and the metal in the bridge? Or something specific about the power cable arrangement - in some cases, the power lines above a railway are arranged with a parallel feeder wire deliberately to reduce electro-magnetic interference, but not always.
However, when using a hand-held compass to check for any magnetic interference, deviations of up to 140° were observed over localised regions of the bridge below which the railway track’s overhead high-voltage wires were being ducted
Sounds like the investigation found the interference to be localised to the overhead cable areas, not the whole bridge.
I’d suggest crap investigation. The road bridge will have enough steel in it to kill off any power cable effect, even if there is any.
Train overhead power cables are AC, too, that pretty much eliminates any constant effect on magnetic field.
Same as @PingSpike says, flown near enough to national grid cables (far higher voltage and current) to know there’s not been any impact.
Don’t have overhead powered trains here - but if there were I’d go and test this “theory”.
You gotta love :
“the pilot should always be ready to switch out of GPS mode into atti(tude)
or manual flight modes to retake control of the UAS if control is lost whilst
in GPS mode.”
Most drones are not switchable, and for those that are most pilots wouldn’t have a clue!
I’m not convinced, and my perspective is as a bridge designer who often works with EMI/EMC professionals, and also with railway OHLE engineers. Plus observations from the several times I’ve taken off from reinforced concrete without any issues. Magnetic flux density is high close to the contact wire and the catenary wire, and if those are close together (because it was cheaper to squeeze them below the bridge when the railway was electrified) and close to the underside of the bridge, I would not be surprised that the effect is measurable above the bridge. I’ll have a think about where I could go and test this, unless any of our readers are near Stoke Gifford themselves!
Sometimes I have, sometimes I haven’t. It’s a really weird one.
Thick sheet steel platform. No problem.
Deck of a car ferry, No Problem. (Didn’t take off, just powered up to test.)
Then you get something that’s almost no steel and it complains.
I’d still put my money on the bridge.
A mild possibility that if, as you suggest and can happen, the contact cable beneath the bridge is really close (old bridges from pre-electrification days - where they lowered the track the absolute minimum, perhaps) there might be an induced current/magnetic field in the structure/REBAR? But - I doubt it. The magnetic field from a straight AC cable isn’t very powerful.
There’s a move afoot in the rail industry right now to look at going back to super-tight clearances in order to reduce the costs of new electrification schemes, so this may not just be a thing of the past.
Let’s hope they apply the same “super-tight tolerances” to timetable adherence. 
After reading the report and then the claims of the manufacturer something just doesn’t add up. As @Pterodactyl points out the magnetic field around a cable carrying AC is very low to that of DC. Because the current is alternating the magnetic field would be collapsing 50% of the time. If the frequency of the AC was at standard 50Hz the test compass needle would reflect this as the magnetic field reverses.
Most likely cause.
The operator powered up the drone on the bridge and received an anomalous compass warning, probably due to the metalwork supporting the bridge. He/she then performed a compass calibration while on the bridge. After calibration everything showed good to go. However after take off the drone was out of the magnetic influence of the bridge, which the compass had been calibrated against, and so the reported heading by the compass was significantly different to that reported by the GPS and bad things happened.
Just an idea.
Nidge.