I'm not sure how sensitive the reservoir routines are to geometry, but I'd make sure that your ESA > PSA, and EVOL > PVOL, since SWAT uses these two known points to infer reservoir geometry (area as a function of volume). I.e., SWAT keeps track of volume by mass balance considerations, and then uses the area to calculate evaporation and seepage losses. I usually set all 12 months of my target volumes (STARG) equal to the principal volume, since I don't want SWAT to be calculating target volumes based on soil moisture.
But your model should be sensitive to NDTARGR, as long as NDTARG > 1. If NDTARG = 1, then the reservoir has no storage and everything leaves the same day it enters. It is unfortunate, but NDTARGR is limited to being an integer in SWAT. It would be better to have it typed as a real number, which would allow much finer calibration of outflow -- and it should be a trivial code change to fix this.
To see the effect, further, your emergency volume must be substantially larger than your principal volume, since the release of water from storage only occurs at volumes intermediate between emergency and target volumes. Everything over the emergency volume gets dumped immediately, the same day it arrives.
And -- make sure you've selected the option to have reservoirs use the targeted release.
Cheers,
-- Jim