Applying also the logarithmic BBSO method to the rainbow images, we get the following results:

We see much more scatter in the early part of the sequence, compared to the ordinary BBSO method. This is both the part where we had the lowest fluxes (shortest exposure times) and (we presume) a proximity to the ‘rainbow phase’ – so are we sure if we see scatter due to rainbowing or just low exposures?

As earlier we expect that scatter rises when erroneous additive levels are removed – but do we have a way to decide if ‘too much’ has been removed? The log method is able to ‘remove too much’ while the linear method is more likely to underestimate how much is to be removed.

The larger total/DS ratios in the early part indicate lower Earthshine intensity. Remember that the rainbow is supposed to be brighter than ‘non-rainbow’ Eartshine.

It is probably realistic to assume that problems have arisen in the early part of the sequence due to low exposure times, not because of low actual fluxes.

Shall we consider co-adding frames into bins – every connected set, for instance?