There is the clock as it exists and there is the clock as it is observed. The existence is the reality, the observation is not. A thing doesn’t have to be observed to exist.
Your last sentence is correct, but it obscures some rather important distinctions. In the context of special relativity, the word "observe" has a rather specific meaning, and it is a
different meaning from what you merely
see. An observation in special relativity is an idealized measurement. It does not
create reality, certainly, but if it is an observation (again, using the term in a rather specific sense) and not just any old measurement, then it
does indicate reality. That's the whole point of using the term. There are all sorts of optical illusions which result from the finite speed of light which apply to what you
see or even measure, but when talking about what you
observe none of those apply: observations
mean what's really going on.
The example I like to use is airplanes flying overhead. Because of the finite speed of sound, you hear noise coming from some point noticeably behind where the plane overhead is. But if you know the speed of sound in air and you know how fast the plane is going, you can calculate when the sound was emitted and where the plane was when it was emitted, and you can
observe that the sound is indeed coming from the plane and not from behind it.
A thing that exists is potentially observable. If we’re talking about human created clock time, then the “correct time” is the “correct speed” that the clock runs at according to it’s design and purpose. In other words, a correctly functioning clock runs at a predictable, constant rate when observed at a local level (in it’s frame). That the clock can be observed to be different than it’s actual existence, from the distance and movement of another frame, doesn’t mean that that it is different.
The
clock is not different: time itself is.
Let me give you an example. Consider just Newtonian physics (with Galilean relativity). You're standing on the ground, and your friend is riding on a train car moving at a uniform speed of 2 m/s. In your reference frame, two events occur, one at x=0, t=0, and another one at x=10 m, t=3 s. Now, what's the distance between these two events? Well, in your reference, it's 10 meters. What about in your friend's reference frame? Well, if we set it up so his origin (x'=0, t'=0) is the same as yours, then the first event happens at his origin, but the second event happens at x'=4 m, t'=3 s, and the distance between the events is only 4 meters. Why the disagreement? Is his ruler broken? Is it changed? No, it isn't. All that happened is that we had to transform our coordinates: x' = f(x). But when I ask the question, what's the distance between the two events in the two different coordinate systems, is the fact that the answers are different somehow artificial? No, the difference is quite real.
So what's different in special relativity? Well, in special relativity, you can't simply say x' = f(x): it's really (x',t') = f(x,t). You need to transform both time and space coordinates. And one of the effects is that the
time between two events (say, two ticks on a clock) isn't the same in different reference frames. Nothing is broken or even changed about the moving clock, other than the fact that it's in a moving reference frame. It's easy to see why this happens with rulers, it only seems strange because from our everyday experience, we don't ever
notice that time has to change for moving frames because the change is so small at non-relativistic frames. But the change is quite real, it is not an artifact, and it's been experimentally confirmed plenty of times.