There are two possibilities here:
1. They are indeed monitoring private messages, which is possible (but unlikely in my opinion).
2. They have set up spamfilters to block certain words or phrases that can be bad or suspicious. When someone triggers such a spamfilter, the network opers are notified of this and can see the message that was sent.
The only way to prevent both 1 and 2 is to start a DCC Chat session with someone instead of a query/private message.
