Both clocks will show 4 pm when they meet. The scenario is completely symmetric as you state it, so whatever applies to one of them also applies to the other. That's the short answer, and the correct one, but of course it's not very satisfying as it stands.

So if you want a longer answer, read on, but be warned: it won't be as simple as the above.
You don't give a reference, so I don't know which "wiki" you are reading (Wikipedia? If so, which specific article?). But just saying that "A's clock runs slower relative to B while B's clock runs slower relative to A" is not really a good way to describe what's going on. For one thing, it leads people to ask obvious questions like the one you are asking.
The thing that is missing from the description you got from the wiki is that it's not just
apparent clock rates that change with relative motion; simultaneity changes too. That is, which events
seem simultaneous to a given observer (which events appear to him to have happened "at the same time") depends on the observer's state of motion. This change in simultaneity has to be *combined* with the
apparent change in clock rates, if you want to predict what someone else's clock is going to read when you meet them.
In your scenario, when A and B are at rest relative to each other, their
sense of simultaneity is the same. So the two events "A's clock reads 3 pm" and "B's clock reads 3 pm" will seem simultaneous to both A and B.
But once A and B start moving towards each other, their
senses of simultaneity are different. So, for example, the event "A's clock reads 3:30 pm" will *not* be simultaneous with the event "B's clock reads 3:30 pm", to either A or B. In fact, A will find that when his clock reads 3:30 pm, B's clock reads something *later* than 3:30 pm "at the same time", according to his sense of simultaneity while he is moving. That means that, according to A, when the two meet, A's clock will have ticked off half an hour from 3:30 pm, but B's clock will have ticked off *less* than half an hour "in the same time"--meaning, from what B's clock read "at the same time" as A's clock read 3:30 pm, according to A.
But how can B's clock read *later* than A's, according to A, when A's clock reads 3:30 pm? Because A's
sense of simultaneity changes as soon as he starts moving--that is, right after his clock reads 3:00 pm. So, for example, when A's clock reads 3:01 pm, according to A's
sense of simultaneity while he is moving, B's clock will read something *later* than 3:01 pm. But B was supposed to start moving when *his* clock read 3:00 pm, just like A; so A will have to conclude that, according to his
sense of simultaneity while he is moving, B started moving *earlier* than 3:00 pm. (That is, earlier by A's clock--the event "B's clock reads 3:00 pm", according to A's
sense of simultaneity while he is moving, happens "at the same time" as A's clock reading something *earlier* than 3:00 pm.)
That's how B's clock can "run slower" according to A, but still read 4:00 pm when A's clock reads 4:00 pm; B's clock runs slower, but B started moving *earlier*, according to A's
sense of simultaneity while he is moving. So if A wants to predict what B's clock will read when he meets, using the frame of reference he is in while he is moving (and in which B's clock "runs slower"), he has to take into account this change in simultaneity. So A will find that B was moving for *longer* than 1 hour, by A's clock--just enough longer that B's clock, which "runs slow" according to A while he is moving, ticks off exactly 1 hour while B is moving. (And of course, B comes to similar conclusions with regard to A, just with everything reversed.)
This post has gotten pretty long, and I should stop and let you digest it.