When deciding on a Yagi antenna, or when figuring out which cell tower is better to connect to, it pays to look at the normalised field pattern or 'radiation' pattern of the antenna in question. When using computer software to determine what's best for your situation you must factor in the antenna type to avoid spurious results.
We're going to demonstrate the difference between the incorrect assumption of an omnidirectional antenna on a cell tower and the actual directional antenna. In this situation the antenna is an Argus CPX209R with a horizontal beamwidth of 33 degrees and a curious looking radiation pattern - specification sheet here. Because we are telling the computer that the transmitted power is 38W (including gain) in both situations, an omnidirectional antenna will have a very overstated performance in most directions.
We can quite clearly see the large lobe (to the north-east), the two smaller side lobes, and the thin rear lobe (south-west). You can imagine if we were far off to the east of the antenna, we would have overestimated the performance we would achieve from this transmitter.
This might be better exemplified by a plot where the transmitter and receiver are 20m off the ground (to adjust for terrain effects). In these examples the antenna is facing 90 degrees from north (or zero on the field pattern).
When the active transmitter is located at a significant height it is absolutely vital that accurate E-plane data (for vertically polarised systems) is employed when generating coverage models. You can see here in the below image the area of reduced signal power due a between-lobe effect.