Does it matter? The refinement will only use observed reflections.
As Andrew says your output from dataprocessing includes a complete list of all possible indices to the upper resolution limit with a Free R flag assigned to the indices. This makes it easier to keep the same FreeR assignments for all your sets of measurements, which is a good thing..
There is a library call which REFMAC could use which would return the actual limits of the observed data - that small modification to the program would give a more sensible log file for the refinement.
At present the REFMAC outputs a reflection list will include all indices, with the Fcalc PHIcalc information listed for missing data . I find that quite useful in some cases, but if you do not wish to use the terms there are ways of excluding them from subsequent calculations..
Eleanor