What would you recommend regarding how to choose two thresholds (LO, HI) for adjusted copy ratio (CR), such that CR < LO is called a deletion and CR > HI is called a duplication?

We use the callAlterations output, with homozygous deletions called as deletion and amplifications called as copy number >= 6 when focal and ploidy < 4 and >=7 otherwise. But these cutoffs are always hand wavy. For some genes, smaller copy number gains can be quite significant (I cannot find the relevant paper, but it was I think a TCGA pancancer analysis). It makes always sense to plot the genes of interest. Sorry that there is no better answer.