Generating a genome index is very slow when the genome doesn't contain one of the nucleotides.

[jiyongkun]

When using meRanGs mkbsidx, the SAindex (k-mer index) generation is taking much more time for C->T or G ->A genome, which, in hindsight, should have been obvious. SAindex contains the positions of all k=14-mers in the suffix array. With one base missing, a lot of k-mers will be mssing, which slows down the calculation significantly.
Please see the STAR issue: alexdobin/STAR#1263