After watching Tommy Edison using an ATM (and how long it took him), my thoughts were mainly stuck in two ideas: he was forced to carry earphones with him and the instructions given were in some way directed to people with no visual impairments.
Although earphones are not objects too big to be carried along, one does not always have a pair in his/her pockets (especially in situations where you don’t have pockets!). Thus, using the ATM becomes something that needs to be planned at at least some level. There is little room for spontaneity.
A solution for this is to provide earphones at the ATM itself. The problem with earphones is that they go right in the ear, so having different people using the same pair does not sound like a good practice. Though probably a bit more expensive, headphones are a better solution. They could be chained to the ATM, or some other option that prevented people from stealing them. In any case, it seems reasonable to have the instructions said only for the user, to avoid bystanders hearing details that only concern the owner of that bank account.
The second thought was on what we base our metaphors. In the video they point out a location by using a clock metaphor. This is based on a visual aspect that a blind person can only learn if explained, but it really does not make that much sense (especially if we can just use a different one). A similar situation happened to a researcher developing a software that taught yoga to blind people: when the metaphor “try to touch the sky” was used, users were a bit at a loss. However, when they were told to try to reach a high self, they completely understand. This shows that, as usual, it is crucial to be able to put ourselves as designers in the shoes of the users and see what kind of instructions fit best in their everyday life.
Non-excluding Design 
Very good thoughts here. The problem with the earphones is a difficult one to solve. Even headphones might cause some hygienic issues. Although better than earplugs, I think the situation needs to be a different one. I have often thought about using the mobile phone, using bluetooth for the connection. Most people, and especially blind people carry a phone most of the time (e.g., to be able to call for help).
Metaphors used are very often based on a comparison with normalcy. “Like a walk in the park” might not be the same for everyone. As you say, the metaphors and directions should be adapted to the user and the context. We fail far too often in doing this properly.
I tried to avoid anything that involved assuming the user had counted on that, as it may prevent them from being spontaneous. Their phones may run out of battery (or they may have one that is not a smartphone). But then again, especially as this is not a life/death situation, I do agree: in this case it sounds better to assume they have a -charged- smartphone than to try to avoid the hygienic issues.