Lepidoptera scales are modified arthropod setae, which are cuticular formations and do not contain any cells or parts of cells. They consist of products of external extracellular secretion, e.g. chitin, selected proteins, wax and others.
See David Lees' paper (2012) on DNA extraction from Cameraria ohridella. David didn't even try scales because the amount of DNA attached to a single scale is just too small. Entire wings work but, as is the widespread practice, hind legs are best even for these mini-leps.
Which brings me to the question - why do you want to extract DNA from scales ? Is it because scales on the eggs would identify the egg laying females ?
The idea was to collect scales from the flowers left behind by pollinating moths, extract DNA and identify pollinating species for each individual flower (as there are multiple co-pollinators in my system). But seeing that this is very unlikely, we have decided to use a different approach that does not involve DNA extraction.
Ah, then perhaps eDNA is an option ? I don't know of anyone doing exactly this but I have friends who are IDing multiple organisms from tiny samples of plants or of water. And, as it's possible to get Denisovan DNA from cave soils, then flowers must be within technological reach don't you think ?
Send me an email if you want to know my eDNA contacts.