All data taken before November 26, 2007, are subject to a small amount of amplitude decorrelation dependent on the difference in delay length between the two antennas in a baseline. The program uvdecor attempts to correct for this:
% uvdecor vis=xxx.mir out=yyy.mir delaymax=8550
Note that the integration times (now baseline based) are adjusted (decreased) to account for the increased noise on baselines with longer antenna delay differences. The value of delaymax=8550 (nanoseconds) was emperically determined from good fringetest data in the 2007 B array, in which the amplitudes dropped linearly with delay differences. The delaymax value is where the amplitude would have dropped to 0!
Especially if your source is extended, it is highly recommended to play with this option for B-array data (with delays up to 6000 ns, decorrelation up to 70%) but even in C-array data (delays up to 2000 ns, decorrelations up to 25%) it should be considered.