Dear Mikkel,
I'm glad you sorted out the main reason your results were lower than you expected.
As you know, the pvwattsfunc module requires both a beam (direct normal) and diffuse horizontal radiation value for each time step. It calculates the resulting POA radiation from those values (and from sun and array angles calculated from the time and location data). I suspect that your assigning the total radiation to the beam variable is causing the difference you describe.
You might want to try using a model like the NREL DISC model (which I discuss in this
post
) to estimate the beam and diffuse components of the total radiation value, and assigning those values to the beam and diffuse inputs to see if you get results closer to what you expect.
Best regards,
Paul.