We have been using the following formula to add reach from month to month for our online campaigns: (R1+R2)-(R1*(R2/100))*.96. After some research on your site, the initial part of the formula seems to be the formula for random probability but I'm not sure where the multiplying it by .96 came from. Any thoughts? Could that be another estimate on duplication?
Applying a factor like "0.96" is an old technique to adjust random duplication (which, as you say, is the first part of your formula) for the fact that duplication in some cases is somewhat greater than simply random.
Between different media, such as print and tv, it is thought to be truly random, that is, there is no greater likelihood that a newspaper reader of your campaign will see your tv campaign, than any other two random events.
However, between two elements of the same medium, like two TV dayparts, there is a more than simply random chance of duplication.
That is the traditional case for using a factor like 0.96.
Between consecutive time periods of the same medium, as in your case, the Guru expectss a much greater chance of duplication. You are looking at new exposures of the same vehicle, which should be represented as accumulating along one sharply flattening asymptotic curve (see below). It's a "cume," not a "combination." Random combination is far too optimistic. Unduplicated users from the first few months to the next added would probably become virtually total unless each month used unique, unrelated sites.