As a further fruitful exercise, it might be worth considering: if the failure rates are not multiplicative, what else could be going on? Here are three alternatives:
1. If enhancer A failed to induce in a given because the whole nucleus of that cell had too little TF activator, enhancer B would also be likely to fail — since the two enhancers share the same activator. If activator is limiting, you might expect the 2 enhancer construct to have a failure rate equal to the smaller the two enhancer’s failure rate. i.e. combine the 0.12 guy and the 0.25 guy and see 0.12 failure.
2. If failure to induction is dominated by blockage of the promoter (e.g. by a positioned nucleosome), such that the 12% failure rate seen with enhancer 1 is the result of 12% of promoters have tightly positioned histones that the enhancer can’t induce to remove, the combined effect of both enhancers would also be simply the minimum of the two.
Here’s another example in the opposite direction, if the more distal enhancer could interact with the more proximal enhancer as a stepping stone to get closer to the promoter, in the two enhancer rate construct it could have a faster rate of interaction and a lower probability of failure, and you might see better than multiplicative. If the 0.25 error-rate distal enhancer can take advantage of the primary enhancer in some cooperative stepping stone like way, then maybe combining the 0.25 with the 0.12 you see 0.99 instead of 0.97.