I am unable to get the procedure described in the MultiWFN manual 4.23.5 to work
I get the error "coordinates do not match"
I am using GJF files as input and submitting the jobs to G16 using Gaussview 6
I am converting the chk files to fch using FORMCHK.
I notice that FORMCHK gives messages "coordinates rotated".
What sequence of operations using G16 should I use?

Many thanks

Please properly use "nosymm" keyword to avoid Gaussia automatically reorientating the system during calculation. This point has already been mentioned in Multiwfn manual.

Many thanks

But if one were to provide the correct extended transition state Fock matrix to Multiwfn (read from a file) then it should give the exact result?

Yes

I'm trying to reproduce the numbers of of ammonia borane from this article:
https://pubs.acs.org/doi/pdf/10.1021/jp209712s
This is the original author of the ETS-NOCV method and it uses the ADF implementation and the  theory level is BP86/TZ2P.
The geometry (of the monomer) is given in the SI.

For the monomer (M) in the Table of the paper we have (BP86/TZ2P level of theory):
dominant orbital interaction: -65.5 kcal/mol
total orbital interaction: -75.8 kcal/mol
total interaction: -44.7 kcal/mol
-----------------------------------------

If I read in Molden files (generated by ORCA using BP86/def2-QZVP) for the dimer and the two monomers (making sure to have reordered the coordinates appropriately) I get with Multiwfn:

dominant orbital interaction: -58.08 kcal/mol
total orbital interaction: -70.170 kcal/mol
total interaction: -45.257 kcal/mol

So the total interaction energy is fairly well reproduced  ~0.5 kcal/mol error (pretty good, considering different Slater vs. GTO basis sets) but the total or individual orbital interaction are off by >5 kcal/mol.

For the ETS step in Multiwfn I tried to read in both the Fock matrix directly (from ORCA outputfile, using -1 option) and also to recalculate the Fock matrix based on the orbital-energies information present in the Molden file (-2 option).

If I use the ETS-NOCV implementation in ORCA instead I am able to reproduce the ADF-values really well:

dominant orbital interaction: -65.827 kcal/mol
total orbital interaction: -76.112 kcal/mol
total interaction: -45.257 kcal/mol (identical to that before)

Total orb interaction is off by only 0.31 kcal/mol and the dominant interaction is off by only 0.33 kcal/mol.

So I suspect a bug in Multiwfn but maybe only for the ETS part, not the NOCV:
The largest eigenvalue for the dominant pair is: -0.557 when using the ORCA implementation
The largest eigenvalue for the dominant pair is: -0.55726 when using the Multiwfn implementation, so identical.

Anyone got any ideas?

Geometry that I used:
B        0.000035390     -0.000019910     -0.125149000
H        1.174850000     -0.000074730     -0.440064000
H       -0.587435000     -1.017410000     -0.440035000
H       -0.587331000      1.017440000     -0.440002000
N       -0.000044690     -0.000077070      1.533410000
H       -0.953218000     -0.000162730      1.904150000
H        0.476727000     -0.825587000      1.903830000
H        0.476593000      0.825494000      1.903860000