I full agree with your statement:

If the line section is not exactly a half wave, then the real loss
factor might be higher or lower depending on the location of the current
and voltage maxima and minima and the relative contribution of R and G
to loss. So, the formula may have significant error for short lines that
are not exactly a half wave.

But I am not certain about this other statement:

Straight away, that tells you that the VSWR must be almost the same at
both ends for it to not matter which end is the observation point, so
therefore the first assumption is that VSWR is approximately equal at
both ends of the half wave.
If a practical line is very long, it cannot qualify as having a constant
VSWR (unless it is 1, in which case the formula is unnecessary), so the
formula is not suited.

I have a feeling that the ARRL chart makes reference to the SWR at the antenna,
and that it DOES take into account that, for a lossy line, the line portions
closer to the transmitter are subjected to a lower SWR.

I try to explain my argument. Let us assume that the line consists of the
cascade of many identical line pieces, each having a 1-dB loss, that one can
freely add or remove. Adding a piece causes an increase of line loss by 1dB +
some extra loss due to SWR. If you add many 1-dB pieces (that corresponds to
increasing the total line loss), the chart shows that the extra loss caused by
the last added pieces gets smaller and smaller (the chart curves all tend to
saturate for an increasing line loss), and this could be explained by the fact
that the ARRL formula does take into account the fact that the last added pieces
are subjected to a lower SWR (and hence yield a lower extra loss) .

What do you think about that?

73

Tony I0JX