One has to be aware, as I'm absolutely sure you are so this may be for the benefit of others, that the % error between 43.50 and 44.49 is smaller than 17.50 and 18.49. As such looking at the figures for larger size wire and calculating down should produce more accurate
For a start, you seem to have got that back to front. Per BS 7671 figures, the 44 (with the smaller potential percentage error due to rounding) id that for 1 mm² cable, but the 18 figure (with the larger potential percentage error due to rounding) is for 2.5 mm² cable - so,
for the comparison of 1 mm² and 2.5 mm² cables, the 'the more accurate' figure (the one with the smaller potential error due to rounding) would be expected to be the one for the
smaller size of conductor (the opposite of what you suggest).
However, it gets more complicated, dependent on what actual CSAs one compares, since, at least in the BS 7671 figures, the degree of rounding varies They don't round to a certain number of decimal places but, rather, to two significant figures.
Hence, for 10mm² cable, the figure is given to one decimal place, namely 4.4 mΩ/m (for two conductors), the potential range is 4.35 to 4.44999, a difference of nearly 0.1, which is about 2.27% of 4.4. However, for 4mm² cable, the BS7671 figure is rounded to 'zero decimal places', namely 11 mΩ/m ( (for two conductors), hence a range of 10.5 to 11.4999, a difference of nearly 1.0, about 9.09% of 11 - hence this time with a smaller potential percentage error with the
larger size of conductor.
I was merely agreeing with the error that you clearly explained exists in other parts of the documentation. However a very quick calculation with pencil so possibly not perfect, the extremities of the rounded figures: 12.1/7.41 do not reach 5/3 so it appears something is not right.
Yes, although the discrepancies are fairly small, it is hard to explain why, from any one source, the ratio of resistances for two different CSAs should not always be exactly the inverse of the ratios of the CSAs.
The actual figures (for any CSA) may well vary a bit from source to source, since they might assume slightly different resistivities, not to mention the fact that different sources quote resistivities (or cable resistances) at different temperatures - but the ratio for two CSAs should still be 'as expected' for a given source.
Kind Regards, John