OK to backport ?

The approach is right, measuring against real PDF width instead of counting characters, and making it independent of MAIN_DISABLE_TRUNC solves the masking concern I raised on the issue. Two problems with the IBAN call sites though, both from the signature change.

The two signatures do not line up:

dol_trunc($string, $size, $trunc='right', $stringencoding='UTF-8', $nodot=0, $display=0)
pdfTruncateText($pdf, $text, $max_width, $ellipsis_pos=1, $ellipsis="…")

and the call was converted positionally:

$this->pdfTruncateText($pdf, $iban, 6, 'right', 'UTF-8', 1)

Traced through your own mapping code, that resolves to:

max_width    = 6
ellipsis_pos = 0          ('right' is mapped to 0)
ellipsis     = 'UTF-8'    <- the encoding argument lands in the ellipsis string

So the ellipsis becomes the literal text "UTF-8", and the trailing 1, which was $nodot in the old call, is silently dropped. With ellipsis_pos 0 the ellipsis is blanked, so it may not surface visually, but the argument is still wrong and would bite if the position ever changed.

Second and more important, 6 no longer means the same thing. In dol_trunc it was six characters, in pdfTruncateText it is six user units of width, so millimetres. Six millimetres fits roughly two characters at a normal font size, not six, so the masked IBAN gets shorter than intended. Worth checking on a generated invoice.

Affected sites, all three the same pattern:

facture/doc/pdf_crabe.modules.php
facture/doc/pdf_octopus.modules.php
facture/doc/pdf_sponge.modules.php

Those three are the masking case rather than layout, and they are the ones I flagged on #40005: they exist to print only the start of a bank account number on a customer-facing invoice. Getting the width wrong there shows fewer characters, which is harmless, but the argument shift is worth fixing regardless so the next person reading it is not misled.

Everything else in the diff, the Project title and RefCustomer conversions, looks right to me since those are genuine layout truncations against a MultiCell width.

OK to backport ?

This looks a structural change that may have a lot of side effect, so it is a change not elligible for backport on past stable versions.
Also option MAIN_DISABLE_TRUNC is not a feature but a dev feature to help to make a transition of this change for future version.