I suspect either the earliest sunrise or the latest sunset would be most useful to late neolithic/ early bronze age people- though I also suspect the winter solstice would be more useful too. And measurement at the equinoxes would be more accurate, as the sun's position is changing faster then. After all, the important thing is to stop the day count/ lunar count getting too far out wrt the seasons, otherwise you'll be celebrating harvest in May or planting too late.

Colin Renfrew's excellent book "Before Civilisation" has some cogent things to say about the relationship between early farming and astronomically- aligned megalithic constructions. One interesting possible precursor he cites is the Hopi horizon calendar. By observing where the sun is on the horizon at significant times of the year, you can ensure that the events belonging to those times take place at the same time every year. This assumes a reasonably open horizon, but with distant hills or other features to act as fiducials, and obviously only works for one community- other communities would have to create their own calendar.

It would be no great step to mark the events with stones aligned with the horizon features, and once you've done that you have a recipe for a transferrable calendar, perhaps associated with a greater scale of social organisation and the emergence of a priestly/ aristocratic power- caste.