What about: SDCCH availability (all SDCCH are working ?) 10% congestion with 10 erlangs only, it looks like only 2 SDCCH/8 are working.
If they are all working, then you could assume that there is a peak of sdcch traffic during 30 minutes in the cell. Since you see the traffic erlang during one hour, it doesn’t look that high. But actually, you have most of the SDCCH traffic in 30 minutes (–> 20% congestion) and less traffic in the next 30 minutes. In average, it gives a congestion of 10%.
It can happen in case of “bursty” traffic, such as when many MS are crossing the boundary at the same time (a train, end of working day, traffic jam, …)
Perhaps you could increase the periodicity of your QoS reports, to verify this assumption ?