The tool splits based on differing read groups. Read groups would need to be added before the BAMs are merged. Most mapping tools include this option now do add at that step in the workflow, or you can use Picard's AddOrReplaceReadGroups to add this in any step prior to when the data is merged.
My guess is that your BAM dataset does not include read groups or that for some reason all are the same. Some modifications to your workflow are likely needed before using this tool.
Hope this helps! Jen, Galaxy team