That's a great question. The short answer is this: The concept that the "other ('moving') clock runs slow" applies only in the case where the observer is in an inertial reference frame (i.e. is not accelerating; or more generally, does not change their state of motion). There's also an effect due to gravity; but let's ignore that for now.
In the experiment you describe, presumably both girls motions are mirror images of each other (i.e. relative to the control clock they both leave at the same time, go the same speed and same distance, turn around at the same time, and arrive back at the same moment.)
Now let's say that each observer is sending out a radio "tick" each second. The other observers can count these ticks, and do a calculation (based on the tick frequency, relative speed, distance, etc.) to figure out the "current time" on the other clocks ("current time" in quotes because that is a relative quantity that not all observers will agree on).
Girl 1 observes the following (based on "tick" calculations):
1. On her (constant speed) outbound leg, when "T" represents the time on her own clock, she finds the control clock is lagging behind T, and Girl 2's clock is lagging still farther behind T. Note this is her conclusion even AFTER she accounts for the travel time of the "tick" signals.
2. At a certain point, she decellerates, stops, and accelerates in the direction back toward the clock. While this is going on, she observes that the control clock is suddenly ticking much FASTER than her own, and in fact it passes up her own clock's time (according to her calculations). She notes that Girl 2's clock does the same thing, though at a different rate. By the time Girl 1 has accelerated back up to speed, she finds that BOTH the control clock and Girl 2's clock are now AHEAD of her own.
3. As she continues at constant speed back toward the control clock, she observes that both the control clock and Girl 2's clock are ticking slower than her own. However, the readings on those clocks are still AHEAD of her own clock's reading.
4. By the time she gets back to the control clock, she finds that the control clock, though it had slowed down, is still reading ahead of her own. Meanwhile, Girl 2's clock has slowed down so much that it now reads the same as her own (Girl 1's) clock.
And Girl 2 records an exactly symmetrical chain of events. All three observers DISAGREE about whose clock was running fast or slow, and whose clock was ahead or behind at any particular instant. But they all agree that, once they were all back together again, the two girls' clocks both had the same reading, which was behing the control clock's reading.