That is very interesting! "My" bands aren't mine, but lifted from a local agency and tweaked very slightly to shrink the number of bands a bit.
The really interesting point is the shape of the two plots which, ignoring the shift for different peak rates, track each other almost exactly, apart from a possibly optimistic blip in "my" rates just post Easter, and the leap for October half term (still don't get that being as high as it is - Windy's "band" seems a more realistic level).
I've pretty much arrived at a model that seems to suit me, using 5 or 6 bands as a base. From there, as has been suggested, I may well choose to apply a few minor tweaks based on previous patterns etc etc. I'll post it up a bit later.
What's encouraging, but not really surprising, is that different approaches produce very similar results. The notion of bands is rigid, but it's a very simple model to use as the start point for setting rates in any given year, so maybe it has its place.