The NEC doesn't impose a derate until you get to 24"
NEC 310.15(B)(3)(a) (essentially replicated for you in PEC 22.214.171.124(b)(2)) does not impose any derating factors for raceways (nipples, really) shorter than 600mm (24"), such as yours:
(a) More than Three Current-Carrying Conductors. Where the
number of current-carrying conductors in a raceway or cable
exceeds three, or where single conductors or multiconductor
cables are installed without maintaining spacing for a continuous length longer than 600 mm (24 in.) and are not installed
in raceways, the allowable ampacity of each conductor shall be
reduced as shown in Table 310.15(B)(3)(a). Each current-carrying conductor of a paralleled set of conductors shall be
counted as a current-carrying conductor.
So, you do not need to worry about this derate for your application, since the raceway routing would need to be quite convoluted to exceed 24".
However, you have two problems here
First off, the normal sequence of events is that the raceway is installed first and wires then pulled through it -- this avoids causing inordinate damage to wire insulation by gouging it with the ends of raceway sections or the edges of boxes. You'll need to very carefully pull the wires back inside the subpanel box for now until you have the raceway installed.
Furthermore, and more importantly, you have a problem here -- surface mounting the subpanel put you in a situation where the raceway needs to make an offset bend ("dog leg" or "S-bend") to get from a knockout in one panel to a knockout in the other. This is especially troublesome for you for two reasons:
- Conduit bends have a quite wide radius, especially for fat conduit like the 3" you are using
- Conduit bodies (including L bodies and pull elbows) must be openable without removing finish materials, just like you cannot drywall over a junction box
Furthermore, that giant 3" nipple is going to cause issues if you flush-mount the subpanel (due to the fact that it will split any stud in the way clean in half). The idea of bringing it in via the front of the main panel doesn't work either, because that'd cause merry heck if the main panel's deadfront had to be removed later for servicing (in fact, I'd call it a violation of NEC 314.22's Exception).