That's very common: for example the venerable HPAK 34401A is 6.5 digits, so on its 1 A AC range it will have a resolution of 1 µA, but the uncertainty even within 24 hours after calibration is 0.1% of reading + 0.04% of range, so for a 1 A input the uncertainty would be 1.4 mA, or 1400 counts.
You should distinguish precision, resolution and accuracy. See
here for a nice, graphical illustration of these concepts. The reason why resolution has value beyond accuracy is that often the linearity and noise floor is much better than the accuracy. So you may not be able to say if a particular cap is 0.998 µF or 1.002 µF, but you might be able to know that a cap that measures 0.9984 µF is larger than one that measures 0.9983 µF, despite not knowing the value of either cap accurately.