I think we'll get 6dB by putting them side by side
This assumption is incorrect. You assume they're in phase, but they're only in phase electrically, placing them any distance apart will cause the actual acoustic waves in free space to be out of phase, this will cause both an increased directional effect and cause nulls to exist at various locations especially with reflections.
Calculating something like this is very difficult due to the large number of environmental variables, the math gets out of had VERY fast. If you really need to measure the SPL at a specific point, purchase an SPL meter that will respond at your frequency of interest, I think you'll be quiet surprised at the different measurements you'll get as you move the SPL meter around relative to the transducer, especially in an enclosed room with obsticals in it.
Using
this calculator, if you punch in 40khz you'll find the wavelength of a 40khz audio wave is 0.8575 centimeters, so the exact center of the transducers will need to be very precisely aligned at a multiple of this value, and the directional/null effect and the shape of the lobes will change with each increase wavelength of separation of the transducers.
I'm attaching a PDF that contains a lot of good information about 40khz, including the radiation patterns of a typical disc based transducer (almost all ultrasonic transducers are piezo disc based)