Great! Thanks for using RESCRIPt!
If you made the reference database yourself, and did not cluster the reference sequences, then percent identity isn't really considered (i.e. the reference db can be used for 100-99% and downward). That is, you can use the reference database regardless of your clustering level. For some further thoughts in this see this post.
The default approach, does not cluster the reference sequence data, it is based on the SILVA NR99 database. This is due to the reasons outlined here. However, you can change the options for rescript get-silva-data to download the full reference database. You could certainly cluster your reference sequences yourself by modifying the rescript dereplicate parameters, say at 97%. Your reference sequence would then "consider" 97% similarity.
-Mike