My research confirms this too. All the 1% types I have found are unique - in
other words there are no 5% types that are also available as 1% and vice
versa. So if 1% types are selected, what happens to all the rejects???
Based on what I've heard, it's likely that those parts which don't hit
one of the special-selection points (e.g. the 1% values) simply get
dropped into the next-less-precise binning category.
I've been told that if you buy a sufficiently-large number of (e.g.)
5% resistors from a reliable vendor, and measure them accurately,
you'll find that they do fall within the 5% accuracy window, but that
you're likely to notice a significant "dip" in the number of parts
which fall within 1% of the marked value. The probability
distribution ends up with two peaks - one above the rated value and
one below. The parts coming through the line which measured out
accurately were selected out and sold as 1% parts; the remainder were
sold as 5% parts.
[I haven't actually tried this experiment myself... consider it
folklore of uncertain heritage.]
By a rough analogy, if you buy "2N3055" power transistors from several
different vendors, you may end up with transistors which all meet the
(very loose) 2N3055 part specs, but which may exceed the specs by very
different amounts in different respects. The "2N3055" parts may very
well be manufactured from dice which were intended for different part
numbers, failed to test out well enough to meet those specs, and were
simply binned as 2N3055s rather than being thrown away.