If the wrap codes are not evenly distributed (i.e 1.5% values are used on 95% of interactions), then when an agent is confronted with a call that does not conform to the norm, he has to decide the most appropriate code for the situation at hand (1 or 2 calls per day). The larger the number of low use items is, the more likely it is that agents will be unable to use them.
For example:
Agent receives 100 calls per day.
10 codes are are used on 95% of calls
5 calls get a code that is one of 690 options.
With 690 options to categorize each of 5 anomalies, chosen by 150 different agents you can expect a high error rate.
We are an inbound C2B shop and have 7 codes. 6 standard codes (Complaint, General Info, Payment, Status, Documents, Other) and 1 that changes based on system issues, current mail campaigns, or natural disasters (or whatever we want to track at the moment). If we start getting a lot of "Other" the Ops team investigates the individual calls to decide if something needs action.
Not really a "Best Practices" answer, but there is some math.