On my line both CRC Errors and FEC Errors are low enough that I find a linear Y axis more readable, so I was unaware of the inconsistency. Having said that, my upstream FEC Errors disappear altogether if I choose a logarithmic Y axis for that graph.
I think G.INP is applied to the downstream more readily than the upstream, so maybe upstream FEC errors will always be an order of magnitude lower than downstream. (Is this correct?)
On balance then, I tend to agree with b*cat, although more from an accuracy of reporting viewpoint than a cosmetic one. However, my logic may be completely wrong - and I probably wouldn't notice if it was changed anyway!