I was able to duplicate your findings and I see what you mean; however, I think the problem lies in what BSmith is designed to do. Instead of deadspace loss, think of spilling a gallon on the floor. BSmith cannot predict that size loss after you've entered a prediction of spilling only 8 ounces. Once you tell it that you spilled a gallon, it can tell you you'd better use more grains and water next time (if you want to fill the carboy.)
BSmith is designed to take the specs and conditions that you have documented on your equipment in the past, and it then does the grunt work math for us, for future batches. Equipment Profile settings do not appear to be interactive like EE% and Batch Size. (Perhaps logically, because it is hard to imagine that the deadspace lost would suddenly increase from 0.75 gallons to 2.0 gallons during a brew.) If it somehow did, then you'd click on "Brewhouse Efficiency" and enter a new Actual Batch Volume of 4.50 gallons, reflecting the additional loss of 1.25 gals. BSmith reflects this as a 16% decrease in efficiency (down to 59% instead of the 75% I used), because it believes the deadspace to be 0.75 gals. IF this condition was expected to persist, then you'd either enter a 2.0 gallon loss in your equipment profile, or you'd enter a 59% efficiency for the next batch and use more grain. Or change your batch volume to 4.50 gals. Either way you have to tell BSmith what to expect and then run with that thru the entire process. (Software-wise, that may be why the equipment choice is made before any ingredients are selected.)
Also a good example of why it is vital to measure all the key stats on each brewing rig in order to get meaningful forecasts from BSmith. All the losses, the evap rate, the batch size, etc., are believed to be correct by BSmith when it calculates OG and everything else.