I do not fully understand your question but determination of offered Erlangs from measured blocking proportions is very error prone. If you explain the problem further will try to suggest better solution or at least explain why this is not advisable.