If this is the same on this fluke meter, they should mention it in the specs/manual.
If it uses an analog TRMS converter like the AD737 (
and the 233 likely uses exactly that actually probably not since it only goes to 1kHz) you end up with a 'residual counts' error with zero input and a correspondingly lesser error as you increase the input to the specified threshold, often 5% of range. This may be a positive or negative residuum and in the negative case, the result is a zero reading before you get to zero. Due to the design and calibration of the AD737 or similar chip, the residual counts have little to do with accuracy above the threshold.
My old F289 had ~200 residual counts, my replacement has zero. Both are perfectly accurate above the specified minimum. As
Fungus correctly points out, you can buy an averaging meter. A quick test of two meters shows that my F289 has a substantial (>5%) error on the 5V range at 1% of scale (50mV) and reads zero at 10mV, but the F27 reads 10mV correctly (10 counts, to the count) and only errs by 1 count at 1mV, meaning it displays 2 when the actual value is 1 count. No analog TRMS setup can do that, AFAIK.
Since the F233 only has a 6A range, that means that the 5% threshold is indeed 300mA and the residual counts error will start to matter below that. I imagine the magnitude of the actual residual error is probably 20-30 counts, meaning some meters might give you random numbers up to 30 or so counts with no inputs and others will go to zero at or below about 30mA input. The important point to remember is that the meter stopped being usably accurate well before you get to the point of zeroing out--if you wanted to measure 50mA with this meter, you'd only get a guess. At 100mA, maybe a good guess.
I forgot to mention, some meters zero out the residual counts in software before you get to zero. This doesn't necessarily mean that they are accurate at every level where they give a reading. RTFM and don't bet too heavily on exceeding the specs by a lot.