Dr. Emma Zahn does a whiz bang job of summing it all up here:.

there really is no such thing as bad publicity.

You citing that ELM study made me think of how those neuroscientists could probably find a lot of "been there done that" fruitful data in marketing and advertising studies of long ago. On the ELM effect specifically, though, I thought right away that if the way the potential customer identifies with the celebrity endorser is related to the product, it might not always necessarily be as they say. I.E., audience has an image of George Foreman as a beefy masculine meat loving type, not one for rabbit food or finger food,  and if he loves how this machine makes his meat, the beefy masculine meat lovers presume they will, too.

By contrast, something like a local sports celeb endorsing a local auto dealer probably works more like the study says, no "logical" reason to take his word for it more than anyone else's when people really engage on making a car buying choice. It's just an advantage of name recognition via that celeb neuron association, that the buyer might check them out rather than not visit at all, doesn't guarantee they will trust them.

I should add that I have a celebrity reaction that I have noticed.  When I see Robert Krulwich's byline or hear his voice on TV or radio, I have this association: the way he does science stories like he's explaining to a five-year-old really really irritates me, I don't know if I want to hear/read this. But there's also another association: it's often on a topic that I would find interesting. So I sometimes will grin and bear him. In this case, it was about a direct interaction between image and brain without any added association by "marketing;" and being in the business of people's visceral reactions to images for a couple decades, that interested me.