I attended a meeting this past weekend where the hot topic of conversation was
exactly your question. We all came to the conclusion that, since RAPD data are
so inconsistent (even using different machines will give you different
results!) it is best to decide what experimental conditions you want to use
(temp, MgCl2, etc) and stick with those conditions for all your RAPDs. If a
certain set of RAPDs does not respond to these conditions, eliminate them from
further analysis. As far as scoring faint bands, I think that may also be an
arbitrary decision on your part. Good luck!