Also, if you use only one source of data (ie. the fat measurement of the scale) and do not mix different sources, accuracy is not that important. Since, if there is a systematic error in the measurements, it applies to all the measurements and the trend is always true, even if there is also a small randomness component.
The main selling point here is convenience. You can do a somewhat decent fat measurement in your bedroom. The Dexa can not beat that.
On the other hand, my model (one I bought in 2015), also supports heart rate and this is total rubbish. To the point that it might give me 120 bpm at rest, then go on and do a workout and immediately after the work out, it reports 70.
I would have said that too, but during the lockdown and stay-at-home period, it made a meaningful rise, and it fell back when I could go back to walking and doing sports. Fat percentage: made up? Probably not. Precise? Also probably not.