The general rule is "If the images don't need to be taken together, don't package them together in a Plan. You are right about the Plan overhead being low. You do pay a price for a pointing update at the beginning of an Observation, but only if it has to come back from somewhere else. If you were running say a survey that takes tons of images, each of a different place, this isn't a problem. But if you have three targets that are somewhat close together, and it bounces around between them, the dilution of observing time could become significant if the images are short.