I agree with avionneta’s concerns.
Even with the relatively low maturity of ACAS, its reliability in terms of avoiding error is better than the human visual system which can be fooled. ACAS may suffer from unnecessary alerts due to system limitations, but as far as I recall it does not suffer from illusions or misjudgement.
A recent example from a busy hold south of LHR on a misty day between cloud layers.
During a turn (looking out) with an aircraft crossing 1000ft below, the other aircraft appeared to be above our altitude and descending. This was due to the lack of a visual horizon and the substitution of our wing axis as the horizon. Thus, the instinctive visual manoeuvre would be to descend, which might then trigger a correct, but potentially confusing RA to climb. The situation could be worse at night where an aircraft’s outline is not visible and any manoeuvre would be flown on lights only – another source of visual error.
For those who would wish to argue that a manoeuvre in misty cloud layer conditions or at night should be flown on instruments, then (a) it is not a visual manoeuvre, and (b) if you are head down then use the ACAS display anyway and always follow the warnings.
Thus the Ops Manual advice is superfluous and could lead to error.