Known issues, in full

The detail behind How This Library Was Made: what is imperfect in the published data, with the codes, so anyone checking can check the same things.

Samples that reflect more than the stock

A transparent ink on the reference stock cannot return more light than the stock. On 8 September 2026 every sample was checked against it. 137 still exceed it and stay as measured, each with its reason in the provenance table:

30 colours were measured on a brighter sheet under a transparent ink and were transposed to the stock — the curve divided by the sheet's gain, estimated from the eight bands where the ink is most transparent. The code is derived from the coordinates, so each carries a new one; the measured curve is archived under the old code, which is retired and never reissued.

8 colours, once on the stock, sat within 1.5 ΔE00 of a colour already measured there and were withdrawn as duplicates:

88 samples whose stored Lab disagrees with their own spectrum

Every sample carries both a reflectance curve and a set of Lab values. Compute the Lab from the curve and it should reproduce the stored one. For 88 of 4253 it differs by 0.05 ΔE₀₀ or more.

The eight largest. The remaining 80 fall between 0.05 and 0.27.

They are not spread evenly. By hue family: 30 yellows, 26 reds, 15 purples, 10 violets, 5 greens, 2 blues. Concentrated in the warm, light, low-chroma part of the space — which is where a small absolute difference produces the largest ΔE₀₀, because the formula weights those regions more heavily.

Why they are left alone. Correcting them would mean changing a published colour on arithmetic rather than a remeasurement. Both numbers are what the campaign recorded.

For scale: the worst, 0.436, is a seventh of the distance between neighbouring colours and fourteen times the instrument’s repeat error of 0.03.

Three samples that fluoresce above the reference stock

0P5.79 Palladium, 0B5.92 Alice Blue and 0P4.90 Blanched Lilac return more fluorescence than the unprinted paper does, which should not be possible if the paper is the only source of it.

Two explanations the data cannot separate: the ink carries a fluorescent pigment of its own, or that sheet was brighter than the reference. Both are ordinary, and the values are published as measured rather than clipped to what the model expects.

More on what the numbers mean: Brighteners: the distribution.

181 colours with no Range score

The nineteen print conditions were computed in April 2024. Colours added or reissued since carry no figure, and are reported as unscored rather than counted as passing or failing — see Ranges: all nineteen.