That is a bit odd. There is a grace period where you’ll continue to receive mlat results even if your receiver hasn’t contributed to results recently, since it’s pretty random whether your data is used for a particular solution or not, but that is not particularly long and doesn’t explain such a long trace. I don’t see a solution near that point in the screenshot that included any of your receivers, the last was 40 mins earlier, so there might be something going wrong there.
The track itself looks correct.
Potentially you can have better range with mlat than with ADS-B, as mlat can be done from a single 56-bit message while an ADS-B position needs a pair of 112-bit messages reasonably close together.