I have worked on a few sites with many cisterns.  One I excavated we took a 
1/4 section of down to the bottom, thought enough to create an adequate 
sample for study (Geismar, et al) and then after washing and sorting what can 
be arbitrary layers of excavation where none are readily apparent, look to 
pieces where they may "crossmend" between arbitrary and/or stratigraphic 
layers.  This helps determine the episode(s) of fill as for example if a piece of a 
dish mends with a piece on the bottom you might consider the whole thing a 
single episode, perhaps part of a building clean-out or an eviction, perhaps.  
Another my friend Nancy Stehling, MS, RPA analyzed was excavated in many, 
many layers with many ceramics, which might give the impression was very 
stratified but may have been done within the constraints of access and time.  I 
recall ceramic shop deposits, with sheets of plywood covered in creamware 
from the "175 Water St. Site" (actual a whole early city block, today filled with 
one building, Geismar et al) that might have resulted from a single episode, the 
shop having the proverbial "bull" run into it, or perhaps breakage in shipping, 
arriving in the port after a rough passage, then tossed out.  I'm sort of glad I 
haven't had to make those decisions in the field!  ANother Civil War era water 
control feature or "cistern" was only excavated by having to stand in it and as 
the joke goes thereby "increasing the sample size".