@richard.siewert in consulting the 2018 Explore Rubric I agree I'd be a bit wary of those choices. Here's my thoughts as I read the different rows.
In Row 1 the rubric defines a computing innovation as "an innovation that includes a computer or program code as an integral part of its functionality." A bluetooth mouse feels more like a peripheral device to me and I think it'd be more challenging to make the case that it meets this standard. The other two feel they might meet this bar but I think the student would need to be careful to describe them such that the connections were obvious.
In Row 6 you can find guidance for how students must describe the data consumed, produced, or transformed by the device. My concerns with a 3D printer or ATM are that it's easy to get stuck thinking about the physical outputs of the device (e.g. money or a print) and the jump to describing the data produced / consumed / transformed by these devices may just be harder for students to make. I do think, however, that you could describe either device in these terms.
In Row 7 you can find guidance for how students should describe data privacy / storage / security concerns associated with the innovation. I think it's probably easier to come up with concerns for the ATM than for the other two innovations.
More broadly speaking, I hear your concerns about students choosing pieces of hardware and based on the criteria above I think they're reasonable. To me it feels like choosing interesting hardware might distract students from thinking about the actual computing happening within the device or the way it is consuming / producing / transforming data. I'd also worry about students choosing to describe the sensors used to collect the data rather than what the actual data itself is (something explicitly called out in the scoring notes of Row 6).
In summary I think there's a bit of risk in choosing hardware as an innovation unless students are thinking carefully about the way the device is consuming or producing data. Beyond that, they should focus on the data itself (e.g. a bitmap photo) rather than the sensors collecting it (e.g. a camera).