The core ARCADY model will give optimistic predictions of the capacity of each individual roundabout of a ring, even if the queues do not interlock. Any interactions between any junctions will have an effect. ARCADY predictions assume that traffic arriving on an approach is free to gain access to the give-way line, thereby being free to take advantage of any gaps which appear in the circulating traffic. Any form of bottleneck on an approach will create “holes” in the arriving traffic, and POTENTIAL gaps in the circulating flow sometimes cannot be taken advantage of; the net result is that a percentage of the POTENTIAL capacity of a give-way line is lost. You need to take some account of these effects in the model, but I’m afraid we cannot quote a figure; it obviously depends on the individual junction.
Fortunately, the Lane Simulation Mode that was introduced in ARCADY 7 and continued through into ARCADY 10 allows you to take account of these “holes” in arriving traffic due to restrictions upstream of the give-way line (in this case another roundabout) and will also take account of the blocking back effects too, if these occur.
As an aside to this question, on the “plus” side, TRL has some anecdotal “evidence” that ring junctions work well, because of its involvement with a trial of one of these junctions many years ago. The first day wasn’t good because drivers didn’t know what they were supposed to do, but subsequently the junction worked very well in terms of capacity.