Marcelo,
I don't understand the (apparent) extra-importance you place on the word "residual", in your replies both to me and to yodon. In my understanding, in the current context "residual" would mean, in everyday language, "actual" or "updated" (risk). It is simply the level of risks that actually (as verified) remains in real presence after the mitigation means have been implemented. If you follow this line, there is not much point in trying to account for (or mitigate) some theoretical risk level that existed earlier; rather, it seems reasonable and practical to further work with the residual risk, ie the one that is actually present.
In that sense I think that the clinical evaluation IS helpful in mitigating such risk, especially when you consider that risk is (at least as I stated) a perceived state of things. Such mitigation doesn't necessarily need to change anything about the device, or any other object or physical aspect of reality. It can, alternatively reduce the level of assessed/evaluated risk by providing new/more information, or new insight into existing information (eg through more analysis). Further, if the Clinical Evaluation includes clinical investigation (eg a clinical trial involving the subject device) it may actively provide information showing the risk to be lower, thus actively mitigating it.
Added in edit: The clinical evaluation will not always be able to mitigate the residual risk, however it should aim at that, and in case of failure (ie when the evaluation concludes that the residual risk is real and correct and can't be practically reduced by means of clinical investigation / assessment) it should be stated clearly in the updated CER. I guess this is why I initially used the term "address" rather than "mitigate" - it is more neutral and doesn't presuppose that risk reduction will always be possible / practicable.