Formula challenge: re-shape and cleanup

mtarler 

I've put thought into removing the 'bellman' records as early as possible. The way I designed this problem (Placing blank rows intermittently) it makes it very difficult to pull those records until before using TOCOL and WRAPROWS. I've considered using MAP to pull the records but the blank rows are an issue (and using MAP to cleanup before TOCOL/WRAPROWS is probably not very efficient).  I had also considered BYROW to identify the rows to be used in the array but again, removing the blank rows first would have to be done.  FILTER appears to be the best option at the moment. Maybe the calculations are a bit quicker if the 'bellman' records are pulled earlier. It would certainly have to be done if the record set was much larger and the 'bellman' records had to be removed to allow HSTACK to run.

My first draft of this exercise actually included arbitrary header information every thousand rows or so (The sheet would've appeared to have been pasted from a poorly designed Word document).  I like creating a puzzle and working towards a solution when it seems impossible. I might revisit that first draft.