Illuminance simulation at constant solar radiation

Hi,

I write because I did some simulations with constant radiation to calculate the illuminance and I got some steps on the results that should not appear.

In the figure you can see constant direct and diffuse radiation for 365 days at noon (12:30)

As you can see there are some bumps that I have discarded as dependant of geometry (Did simulations with different geometries even with just an horizontal surface to analyse and nothing else.) Also discarded from the solar position since the bumps happen in different days as you may see in the graph. I have also tested it with Diva and honeybee so I guess it has to be with RADIANCE. I got this data reading the illuminance files from an annual daylight simulation.

I guess it has something to be with how RADIANCE transform radiation to luminance (luminance = (RED*.263 + GREEN*.655 + BLUE*.082) * 179) And how it determines the values Red, green and blue of the spectrum for a determined solar radiation. I guess, because I was unable to learn how RADIANCE calculates this spectrum, that it has to be dependant of the radiation and solar position with a quite large difference of spectrum. 

Normally this bumps are hiden behing the variation of radiation over the year, and I guess that the tool was not designed to have such unrealistic constant radiation, but I would like to know why this happens.

Can anyone help? Any clue is welcome!

Thank you.