I would set the light at a steady angle (I was thinking 45*, but it might depend on the geography of the area what gives the most visibility), and rotate it. Depending on resources, it may be steadily rotated, or spun around every time it's turned on. Periodically, the colony can use Morse code or something similar to announce their intentions, and the exact angle the light is set to - not while the light is rotating, that would make it hard to pick out precise light fluctuations from light movement, but once aimed in each cardinal direction and possibly at each midpoint as well.
What this should do... First, the angle should improve visibility in the direction it's aimed at. It has to reach sideways to be visible - if the beam is straight up, it depends on waste light (from intensity) to be seen from a sideways angle. It also has more chances for the light to hit clouds, fog, or dust, for the light to bounce off of. Turning the light means a colony in any direction will have the best chance to see the beam.
It will also give a crude directional triangulation. When the beam is pointed directly at (or away) from the second colony, it is a point of light. At all other times, it will be an angled beam, slanting back to it's point on the horizon. Twice during its rotation, the beam will be at right angles to the second colony, and can be visible along its length (under some conditions) tracing back to its origin. It is giving itself as a reference point, since it is equally distant to any points that divide its rotation - ie, it is halfway between the points it hits 5 min before and 5 min after the time when it's pointing directly at the second colony. For any time or any points that fall in that 180*.
Finally, distance can be calculated if the geography allows. Given the angles (via morse code), it is possible to find an idea of the distance given a chance to compare any two sides - pythagorean theorem. If the light hits a cliff or mountain, for example, it is possible to judge the height (roughly). This can be compared to the apparent distance from the origin to the height measurement and give some idea of how long the last leg of the triangle is. If the height or the distance to whatever was measuring the height is known, it can be quite accurate to pinpoint the distance to the colony originating the signal. Additionally, if its height can't be accurately measured, it can still be crudely estimated that the higher the beam looks, the farther it's origin.