The AGCH congestion is different from the SDCCH congestion, or SDCCH blocking. We are talking about 2 different channels !
It is something that is usually overlooked in the KPIs (except for erlang.com addicts 🙂 )
your SDCCH assignment success rate is low, and it COULD be caused by AGCH blocking.
If AGCH block reserved = 1 and there is no dynamic preemption, then you surely suffer from AGCH congestion.
Now, you actually don’t have a problem, 89% is totally acceptable, especially in low-traffic cells in rural areas. Lots of Ghost RACH, some coverage holes. Some RACH Requests might reach the BTS, but the answer from BTS to MS is never received by MS. That counts as a SDCCH assign failure.
You can reduce it by increasing your RXLEV ACCESS MIN from -110dBm to -100dBm, for example (just to check the impact).