This is one of those questions where the answer is YES (en principe :-)
I like to add an "in practice" to such answers.
In practice, statistical errors are usually less important than systematic
errors, and you can obtain a good refinement with what appear to be
surprisingly poor statistics.
In practice you should be using a position sensitive detector, or some
other kind of multidetector anyway, and collecting the whole pattern
simultaneously. (And if you can't afford a PSD, you probably can afford
to spend the weekend collecting data :-)
Variable counting time is just one more thing to get wrong :-)
Alan Hewat, ILL Grenoble, FRANCE fax (33) 4.76.48.39.06 hewat@ill.fr
ftp://ftp.ill.fr/pub/dif http://www.ill.fr/dif/