Much more accurately: "Interesting BRI's efficacy is only in the higher cisplatin dose." If you're taking results at face value, the trial shows that B causes or exacerbates SOM at the lower dose (which is of course unlikely).
For the sub group results to be compelling, the company needs to try to explain why one arm performed and the other did not. So your hypothesizing was at least on the right track.
If the company doesn't acknowledge and explain, then we get the stink of "data mining" (aka cherry picking), and this management appear had enough credibility issues. Dumbed down: It's awesome B appears to work so well just for blondes, but why would that actually be? And why would non-blonde results be actually worse than (or same as) placebo?
We still have explaining to do. The company didn't handle it well.