My guesses are that you are seeing the effects of a stochastic algorithm, mixed with relaxed simulation parameters (low sampling density, few ambient bounces, and relaxed subsampling thresholds). Here's an old-school frown: =(
A couple of things to try here:
- Use the "fine" setting for the Radiance simulation parameters; this will help smooth things out and will probably manage to find more light deeper in your spaces.
- It appears that your illuminance map definitions extend to the edges of the building; offset the map boundaries one or two feet to the interior of the thermal zones. This will pay dividends in a couple of ways. When you calculate daylight metrics and include points right up against the wall, you are hamstringing yourself unnecessarily. Even the IES guidance for calculating spatial daylight autonomy allows you this "wall offset". Including those dark corners in your DA and other calcs is unfair. Secondly, by moving the points away from that infinitely thin wall polygon you reduce the chances for random exterior calculated values to influence your interior ones. This second issue is less likely to surface as long as you take care of item one and increase the simulation parameters.
It's hard to comment beyond that for now without better overall views of the building, but I'm guessing the maps will make a bit more sense when you make the changes mentioned here. Let us know!
