I have looked at the UV and IR light distribution for the following galaxies: UGCA320 UGC8201 UGC2716
(for the other two, the UV data are missing from our master archive).  In all cases, the FUV and NUV light is far more extended than the IR (24 and 70) and, where available, the H-alpha light.

The red UV colors, relative to the FIR/UV ratio, are possibly the result of two combined effects:  when deriving the FIR/FUV ratio, the two fluxes pertain to different distributions within each galaxy; for two of the three galaxies the IR and H-alpha emission regions are more centrally concentrated than the UV emission (for the third galaxy UGC8201, I could not identify regions of IR emission, but the little wisp of H-alpha emission is concentrated in a small region to the right of the image).  Thus, the FIR/FUV ratio will be artificially lowered.

The FUV-NUV color is consistent with the overall absence of widespread H-alpha emission, thus old/ageing UV-emitting regions.

All in all, the location of those galaxies in the IRX-beta diagram is not surprising: they seem consistent with most of the UV emission not being accompanied by ionized gas emission (i.e, they are `older' regions' than about 10 Myr).

This may be the case for many of the LVL galaxies, if one takes into account that these are mainly dwarf, run-of-the-mill galaxies, with not much energetic SF going on.

I also looked at the extinction dependence of the EW(H-alpha).  The full expression is: EW(H-alpha) ~ 10^(-0.4 * 0.57 * A_V,star)

For A_V,star=3 mag, this corresponds to a change in EW of a factor 5, smaller than the trend you observe.  I suspect, the extinction dependence of the EW may account for the spread at fixed perpendicular distance, but not the overall trend of Figure 13, bottom panel.

I still find the fact that the LVL gals are located mostly to the right of the Cortese line a bit disturbing.  The absence of galaxies with a better correlation between the location of the UV and the IR emission may tell us something specific about the LVL galaxies.  Perhaps a selection effect.  Or maybe it could have to do with the steeper IMF (or lower mass stellar clusters) that seems to characterize the lower  luminosity galaxies (Meurer et al. 2009, astroph/0902.0384; Pflamm-ALtenburg et al. 2009, astroph/0901.4335).  One way to test this is to check whether there is a trend for fainter (in R-band) galaxies to be located at larger perpendicular distances than brighter galaxies.  Do you have the data for this test?

--Thanks for this latest round of comments.  The discovery that many times the UV is more extended helps, and I've added discussion to the text on how global IR/UV ratios are artificially lowered and FUV/NUV are artifically reddened.  I've yet to look into the R-band issue.