Our vendor is still experimenting in the lab so no final conclusion, but till this moment they found that this problem happens only with 2 phone manufacturers phones and they think that for particular AMR coding scheme MS indicates that it is sending traffic over TCH, but BS of corse doesnt receive any so it thinks that BER is very high.
With your strategy there are also some drawbacks:
You have to very accuratly use CRO because usually 900 coverage area is greater than 1800 and if you force MS to use 1800 all the time then MS in areas where 1800 coverage is very bad but 900 good will have problems.
FDR doesnt allways help so for your 1800 sectors CSSR statistics probably is pretty poor (in our network for every 20th MS that makes call in congested 1800 before he made cell reselection back to 900, FDR doesnt help, we have only 30-40 calls a day per congested 1800 cell so this is not big problem for us 🙂 )
Now about our drawbacks 🙂
Yes, our congestion, TCH blocking and Handocer success rate statistics looks very bad, but it doesnt affect customers so we arent bothering about most of them 🙂
Yes, there is also a TCH waste in 900 cells, but its small.
Also there is a benefit: we are traying to use 2 TRX per 900 cell, one for data traffic, other for calls and all exspansion, if possible, do in 1800 cell so we have greater coverage with smaller number of sites (different tilts and 3dBm makes all the difference :)) and its very helpfull in oasis like places.(in cities there is different story and aproach)
So our frequency plan is tight but not too much 🙂
SD assignment statistics for us are good and SD drops are over 0.15 only in 1800 because of very high congestion but we will try somehow to reduce it (althought doesnt have a clue how, yet)