You can use -q:a, for "true" VBR, but its metrics are a few percent (imperceptable, we still win) less.
"The benchmarks I posted were done mainly on 44.1Khz. I tuned by ear on 48Khz data though, so some of the windowing/transient logic is tied to 48Khz. It translated to 44.1Khz well enough that I left it as-is, since the timing difference isn't that large."