When this keyword is present and the calculation performed is a FOM-weighted synthesis,
then all reflections with a figure of merit less than f will enter the calculation
with only amplitude restraints, but no phase restraints. The idea behind it, is that due
to the algorithm used by GraphEnt (discussed in page 14), reflections with
a high
F/(F) but low FOM, will effectively be treated as weak. By removing the
phase restraints, strong reflections (but with largely unknown phase) will be treated
as strong reflections with unknown phase. The hope is that their phases will converge
to a value that maximises the configurational entropy of the map (while being consistent
with their amplitudes). Now, this whole thing (of arbitrarily selecting a FOM cutoff
to switch-off phase restraints) is clearly an ad hockery which is neither
promising, nor convincing (but you can give it a try just in case). This is more
extensively discussed in section 16.8. I should add (just to avoid any
misunderstandings), that the iteration is started using the phase angles given to the program
( ie. it is not a random-phase seeded calculation), and that the FOMs given are not
ignored but are used to adjust how fast the corresponding amplitudes will approach
convergence.