I have been thinking about this for a while, and I feel process mining is useful mainly for establishing end-to-end visibility of a process, especially for people sitting at the top who want to view how the process is behaving. But the people who designed the process and the operations team executing it already know how the process behaves, where the bottlenecks are, and the ins and outs of the process. The PI analyst is usually an outsider who can generate a lot of improvement opportunities from the data, but a lot of these can be theoretical and can be rejected by the operations team because they are simply not possible. In this sense, the PI analyst can sometimes be similar to an uninformed management, because they have the information from the process but not necessarily the context of operating it.
Also, every process comes with an agreement to fulfil its instances by the operations team through targets such as SLA and RPH, and the operations team is almost always managing to complete these targets. When degradation does happen, it is often because of a high-volume influx and a capacity gap, rather than something fundamentally unknown about the process. This makes me think that the person best suited to work on optimization would be someone who is the operator + process expert + analyst—someone who understands the process from the inside, knows the operational constraints, and can also analyse the data.