(1) you will not want to allow too many Ns in your reads, especially at 95% clustering threshold, as it can lead to identical read copies not clustering together due to N differences. Since your reads are quite long you could probably increase this from its default value.
(2) With 250bp reads I expect that you will have some reads trimmed to shorter lengths due to decreased quality at the 3' end (if you have filtering turned on; e.g., "filter_adapters = 1"). The default minimum length is 35. If you want to enforce a longer minimum length you can increase it here.
(3) Similarly, if you have low depth data you may have many sites with poor base calls (Ns) and these will affect across-sample clustering, so you can limit the max number of Ns here. NB: this only counts internal Ns in consensus reads, since terminal Ns will be trimmed off.
(4) A high number of heterozygous sites within a consensus read may be a sign that you clustered paralogs or repetitive regions together. Since your reads are quite long you may want to increase this from the default setting.