* "randomi" in your original command does not influence your results in the graph, the following should make your graph beautiful. You can also make your own adjustment based on this.
metan oddsratio lowercl uppercl, label(namevar=study) ///
by(subgroup) nosubgroup nohet nowt nooverall null (1) xlabel (0.8, 2.8) texts(160) ///
effect(Odds Ratio) force