First, for practical purposes, I wouldn't disagree with anything said so far!
Taking the 1:500,000 chart for Southern England, which is Conformal Conical Lambert, and knowing the scale varies between the standard parallels, I wondered by how much, so I measured 1 degree distances between 50-51, 51-52, and 52-53 degrees of latitude.
There was essentially no difference, it was 59.7 nm in each case. But this was "the wrong answer". Not a big deal in real life, but it could matter in an exam.
I suspected my ruler. So I measured my 18" nav ruler with a long ordinary ruler, and 90nm on the nav was 333mm on the ordinary, as best as I could measure. Now 90 x 1852 / 500,000 = .33336m = 333.36mm, so my ruler was OK.
I measured the scale on the chart, and that was also 59.7 nm. So the whole chart was "wrong". I imagine it has shrunk since I bought it, or perhaps it was always too small.
For completeness, I made a spreadsheet to calculate the theoretical scale variation across the chart, and the theoretical distance was 60nm +/- 0.05 nm, which is not measurable.
So the moral? The most accurate way to measure distances on a chart is indeed to compare a distance on the chart against a scale on the chart (either against the main scale or against minutes on a nearby line of meridian), but not for the "official" reason.
For this chart at least, the reason has nothing to do with the theoretical scale variation across the chart, it is because of shrinkage!
Another wasted morning...