As noted above, the individual approach is less automated and can produce fundamentally poorer deconvolutions. However in some high-dynamic range applications, it may be preferable. The reason for this is that the deconvolution process does not depend on the primary beam model (nor do some errors such as a constant pointing error affect the deconvolution). With the `individual approach', it is possible to deconvolve sources (and thus eliminate their sidelobes - which is the important issue) beyond the limit where Miriad's primary beam model gives up. To do this, however, you must make images larger than the main primary beam lobe (see Table 21.2 above).
In the following, we assume that the `joint approach' section has been read and understood.