I expect there to be a lot of error in MLAT positions, but I would expect ADS-B positions to be more accurate than this. I don’t think I’ve seen one this far off before, unless it was jumping around. BTW, I modified the colors so that it is dark brown when on the ground. The plane’s position shows landed and turned off the runway.
http://victorspictures.com/img/s2/v51/p915324889-5.jpg
It’s probably using an inertial position source, not GPS.
One of the parameters transmitted is an uncertainty category that tells you something about the accuracy of the position (NUCp / NIC, depending on version); you could plot that if you’re interested.
Sounds like dead reckoning. So I guess my assumption that they always use GPS is incorrect.
Yeah, I’ve considered modifying the line graphing to spread the line out over the uncertainty. A project for another day. 
Yes, inertial navigation is a dead reckoning system, so it accumulates error over time.
A challenge in trying to plot the uncertainty is that the reported value errs on the side of caution and there’s no upper limit in the worst category so you may end up with such large error bars that it’s useless.
ADSB Output requires a WAAS GPS source (WAAS typically provides accuracy within 1 meter). I would attribute this to a chance glitch more than anything.
Perhaps that is the theory, but in the wild transponders can and do switch to different position sources and continue to send output (with an appropriately degraded NUCp).
The classic example of this is “spiky” tracks where the transponder is switching between GPS and a dead-reckoning source frequently.
There are also plenty of installs which send position info with a NUCp of zero, which I believe usually happens with an install that has not been certified as being suitable for separation purposes. In that case all bets are off about what the position source is. But the data is still there.
Maybe, at least in the US, non-WAAS-GPS is sometimes used because it’s not actually required yet?
Does a Mode-S or S/ES transponder require GPS/GNSS?
WAAS is only available in the North America and Hawaii. There are similar SBAS systems like EGNOS (Europe) and MSAS and a few under development. Most will be redundant/complimentary when Galileo is commissioned.
I looked into it further and Obj is right, WAAS (or similar) is the preferred input for ADSB setup in the USA, however other inputs are acceptable. Per the regs the output only needs an accuracy of NACp 0.05NM, which is 92 meters or so.
Watching older Boeing 767 and Boeing 737 (old generation) land at EGLL frequently showed this error a few years ago - for example landing several miles north of the runway!
Clearly not using GPS to feed into ADSB position information. Not had chance to see whether system updates have improved this.