How so? Are the points often grossly inaccurate (out by 10s of km), are the timestamps invalid, do they simply make up data? If you can tell me why the data shouldn't be relied upon (or at least believed), I'm willing to listen.
The reason I'm pushing FlightRadar24 is because the data is there. I also am "trusting" that their data is real because they show gaps in the data as opposed to filling estimated positions in the track between known points. This means that people are less likely to jump to conclusions because they don't realise they're looking at an estimated great circle track.
I've seen no kinks in the tracks to imply that the data is jumping around, or that the points are collected out of order. What I've read about ADS-B gives me no reason to think that the positions and data shouldn't be accurate - otherwise Australia, the US and EU wouldn't be requiring it for some aircraft.
So we're left with the receiving stations - is it possible that amateur receivers are pushing corrupt data to the website, though from what I've read:
ADS-B messages (transmitted every sec without interrogation, with (plane symbol) or without (ball symbol) position, Mode-S downlink format=17) are encoded together with an unambigious 24-bit CRC checksum and would be discarded by a proper decoder if that checksum is false
so receivers shouldn't be doing that.
So, as far as I can tell, we're left with purposely corrupted data, which I'd like to think wouldn't happen, or if it did would be picked up by users.