Alternative agnsed + pexmon model configurations for the LF data set. Here, we include the median-combined and host galaxy starlight-subtracted UVOT photometry; see text of Section 4.2 for details. We display the unfolded Swift UVOT, XMM–Newton pn, and NuSTAR FPMA spectra; data from XMM–Newton MOS and NuSTAR FPMB are included in the modelling but not shown. Top left: We include both an outer disc, warm Comptonization region, and hot Comptonization region, requiring only that |$R_\mathrm{warm}\lt 10^3$|  |$r_g$|⁠. Top right: Here, we apply a two-component agnsed model, with no warm Comptonization region. Bottom left: Here, we constrain the hot Comptonization region to a compact size, with |$R_\mathrm{hot}\equiv 10r_g$|⁠. This yields a very poor fit even to the X-ray spectra, due to the trade-off between matching the UV and the X-ray flux levels while diverting only a small fraction of the accretion energy into the hot corona. Bottom right: Here, we set the hot region of agnsed to an electron temperature of 50 keV, as opposed to 300 keV in the other cases. We note that the LF UV-optical data are not satisfactorily modelled by any agnsed variant; however, the UV flux level is high enough to allow a strong contribution from the warm-Comptonized emission component identified in our X-ray modelling.