Root Locus looks random

When I plot the root loci of higher-order systems I get random-looking lines instead of the correct smooth branches between the open-loop poles and zeros. The rlocus routine appears to be unable to keep track of which branch a set of roots belong to. Any suggestions?

[James Wiken]

Could you provide an example of this behavior? A root locus plot with its associated transfer function would be very helpful in answering your question.