@sarith can write about this more elegantly but here is some details that you need to know about Daysim (which is used for annual daylight modeling in the Honeybee legacy) versus Radiance which is used for point-in-time daylight modeling.
-
The sky in annual calculation is divided into patches while the sky that is used for point-in-time calculation is continuous.
-
Radiance sky uses the accurate position of the sun for calculating direct sun contribution while Daysim uses a simplified method which uses the closest sun from 61-65 sun positions.
See this reply for more details: HB+, legacy and DAYSIM results - #4 by mostapha
The issue that you’re bringing us here is basically why we implemented radiance-based annual daylight simulation in Honeybee[+]. You can read more here about the differences on this Wiki page:
In short your point-in-time simulation is the accurate one and the annual simulation underestimates the direct sun contribution.