Variable-length Figure Cipher between Henry de la Tour and Duke of Nevers (1589, 1591)

Speaking of variable-length figure ciphers mentioned recently, BnF fr.3619 (catalogue), f.103 (DECODE 9442), is a letter of Henry de la Tour (viscount of Turennes), to the Duke of Nevers dated 29 November 1591. For the most part, the ciphertext consists of Arabic figures.
The cipher used is no.23 (f.46-47) of the Nevers Collection (BnF fr.3995), in which the letters A-I are assigned single digits and M-U are assigned two-digit figures (with a dot over the digit in the tens place). Although the cipher includes instructions to write figures continuously ("Aussi fault aduertir de ne separer les mots come lon fait comunem[ent] ains escrire tout aulong le chifre sans aucuns distinction"), chuncks of two digits are barely visible in the ciphertext.
De la Tour joined the Protestant party of Henry of Navarre in 1576 (Wikipedia). At first, I thought use of Arabic figures came from Navarre (see "A Cipher of Henry of Navarre before Accession to the French Throne (1587)"), but in view of the 1571 instance, it may be more natural to think it came from the Duke of Nevers at least for this case.