the -G parameter refers to intron size (" -G NUM max intron length (effective with -xsplice; changing -r) [200k]") for the minimap2 spliced mapping mode so this would probably really only apply to RNA-seq.

for general whole genome sequencing type data, I think this would be unneeded. also note that filtering using -f 0x800 would completely miss the PRIMARY part of the chimeric read. with chimeric reads (or what i call "split alignments") there is a primary part of the alignment which does NOT get the 2048/0x800 flag, and then all the supplementary (split) parts that DO get the 2048/0x800 flag.

you can filter all reads that have an SA: tag to get both the primary and split parts of the alignment. you can just samtools view yourfile.bam | grep "SA:" as a coarse method to get all splits. i describe some aspects of chimeric/split alignments in my wordy blog post here https://cmdcolin.github.io/posts/2022-02-06-sv-sam#what-are-splitsupplementarychimeric-alignments

and regarding the OP question, default minimap2 parameters should obtain chimeric/split alignments if there are any. there could be parameters that can be tuned but is not NEEDED to get them