Attackers are not interested in her, but rather those associated with her. It could be YOU that is the target, and she is solely being used to get to you. This is how impersonation works on the social engineering level. When I perform social engineering during penetration testing, I impersonate someone with the most solid connections to my target. For example, if my goal is to call a help-desk for a password reset, I wouldn't want to call them pretending to be a technical person, I'd pretend to be someone non-tech savvy, so I would choose perhaps a sales person - older - and I would fill the conversation with so much fluff (deflection).
So you now ask: "Why my grandmother?" Why not? Who (outside of people in the security realm) would stop to think twice about grandma, or an older individual linking up, sending a message, etc. Most would just accept, accept, accept. It's what we (most of us) do as humans, sympathize/try to be friendly.
So how do you minimize this? You can continue filing complaints, but there is little you can do to get someone to stop it from happening. Complain on Facebook enough, might minimize that attack surface, until they sprout up on say LinkedIn, or Twitter, or some other form of social media. What you CAN do if you have "the gift of gab" psychologically, is play along with it, and see what the individual wants, and use that as a warning to other family members/friends: "So this scamster pretended to be grandma. I went along with it, and he/she asked for a wire transfer... Or information about my current mortgage." Your grandmother is not the target, someone else is.