The problem is specifically that ARP/wARP *doesn't* support the IUCr convention as given in IT (Vol. A, >= 1983 edition, Table 9.3.4.1, p.758 in 5th ed.) regarding choice of cell in primitive orthorhombic space groups, and I suspect in centred monoclinic ones also. AFAIK ARP/wARP and pointless are the only two CCP4 programs that currently don't fully support the IUCr convention
Phaser doesn't "support" the IUCr convention, and if it was used for the original MR in this case (I don't know whether it was or not), then it would have caused the "problem". We have had user requests to change the output to the IUCr convention, but other people get confused if the axes are permuted. So the choice will be made an output option - Frank von Delft suggested the keyword "IUCR [ON/OFF]"! Vote for your choice of default now...
Airlie McCoy