Something like this?

Something along those lines was one of the ideas I was playing around with but with finer blocks of time.

select hg from readings where time > "04:00" and time < "4:30";
.
.
. and so on
[download]

and then averaging the time slots. Where I have no values in a given time slot I could to linear interpolation. Abigail makes a very valid point that I am trying to wrap my head around to figure out how to implement it is the theory (fact?) that there may be a sinusoidal relationship to consider as well.

I beginning to think that I have a good basis for a posting to CUFP when I get this working.