Unfortunately, it is not just about the given specs that will determine the output of this idea. You have it sort of correct when you say they may couple and average, but not all parameters will have a large leniency in volume and port characteristics, so in a sense, the modeling of it may be quite inaccurate, or not far off from a single one being modeled in the same air space. The problem you encounter in some cases is each drivers ability to provide specific frequency limitations for each tone played. What this says is that there will be frequencies where the accuracy of a calculated model will be null due to phase issues and db levels due to mechanical specs from each driver. The normal increased output from adding a second driver is the common known 3db audible increase, but this may vary depending on the specific phase of each driver and when they are in phase and out of phase and by how much. This can cause small dips and peaks in the entire response that if you have the right parameters and design specs, will be audible. The ones that are inaudible will have an increased amount of low frequency distortion and can possibly, under extreme circumstances, cause a wobble effect depending on the time delay of each drivers contributing response characteristics.
Most effects of this will be inaudible to the common ear, especially when the specs are not far off from each other, but it is unlikely that every specification will be nearly identical, though possible. That being said, I think others will agree that two separate drivers of different parameters are not recommended for those reasons.
One of the possible noticable effects will be tuning characteristics shown through excursion movement. You may notice, if the parameters are off enough from each other, that during the calculated tuning, one will be on key, and one may be off a bit for minimal xmax at resonance in a conventional design and even in tlines. This can be bad because if excursion control is not equal, the ability to determine cutoff is null and the c,calculation of max power applied will limit the in tuned driver to control the xmax limitations of the out of tune driver so not to damage it. So, a possible Loss on output may be needed to operate both under usable conditions.
These, of course, are extreme measures, but possible and too much of a risk when dealing with different specs needing different parameters of volume and area to control response, phase, xmax, power handling, impedance, etc.
Not something I would calculate for though it is possible through testing. And I have done a bit of testing on it already to find a lot of inaccuracies in the calculations to achieve a given response. This was done with two drivers of even the same model with one broken in and one not....with identical results.
Bottom line.....not worth it.