Hi @barbongiuseppe, this is my first time trying to answer questions on this forum so please let me know if there is any misunderstanding of your question. The content below is based on my understanding and many others on this forum may as well correct me or weigh in with better answers.
Since you are using HB, the annual daylight study runs via Daysim. Daysim uses 61-65 sun positions for direct calculation. Here is a post and Sarith mentioned the Daysim sun position and its inaccuracy on post 5. HB+, legacy and DAYSIM results
In HB annual run, the sun position is taken from Daysim, and the irradiation is taken from EPW file. But in HB point-in-time (grid-based analysis, climate based sky), the sun position is calculated via Radiance by your input (the hour), and the irradiation value is still taken from the same EPW file. So the sun position difference may lead to the difference of illuminance lux values for the same time from 2 methods.
Also a similar post on this issue can be found here: Illuminance Annual vs Grid data
Hope that helps!
Best,
Xiufang