Suppose I see a rabbit for the first time. I know it is small, brown, has four legs, is furry and has big ears. In recognizing the bunny, I have created a filter in my brain - a rabbit recognizer. Anything that is recognized by this filter I will call a "rabbit". This filter is an abstraction because it will recognize any rabbit, including rabbits I have yet to see.
So when I say "rabbits have long ears" I really mean that "my rabbit filter is triggered by (among other things) long ears."
This exposes an important fact about language. It means that when we speak in terms of abstractions, we don't have to be referring to some Platonic ideal or some floating universal. We can be referring to our own faculties, and what would trigger those faculties to recognition.
So when I recognize what a watch is doing when it is keeping time, I automatically create an abstraction filter for time-keeping. I can speak of time-keeping mechanisms in the abstract because I refer to the filter in my mind that recognizes such things. And I can say that the time-keeping ability of this particular watch is due to the mechanical mechanism inside of it.
So when you ask "Is a watch a time-keeper in the absence of minds?" you have to decide what you mean by the question. "Time-keeper" could mean that I presently see and recognize and use the device as a time-keeper. Or "time-keeper" could mean that, if I had such a device here and now, I would recognize it as a time-keeper. You would have to take time-keeper only in the strict, former sense of the word to say that time-keepers would not exist without us. However, taking the word "time-keeper" in this sense is misleading. If I used the first definition, then any watch not in my presence would not be a time-keeper. (And any rabbit yet to be born would not be a rabbit, etc.) No one takes language to mean this. The language is taken such that a device is a time-keeper if it would be recognized as such by a mind, if a mind were present.