My findings are similar to what some others have reported.
I tested TUBE on the latest version 0.33b5 on two different identical rigs, each with 6 x 8GB RX 580, with the only difference being one is on Adrenalin 18.6.1 drivers and the other is 18.5.1.
With each rig, only one GPU out of 6 reaches maximum hashrate (approx 1220 h/s), and the other 5 GPUs level out around 990 - 995 h/s. I have tried various multihash settings as well as other greeks but this issue remains regardless.
If I use the --no-warmup switch, 3 (sometimes 4) of 6 GPUs (for both rigs) will reach maximum hashrate (up to 1180 h/s) and the others will remain in the upper 900s h/s.
How long are you waiting for max hashrate? I had found even w/ previous versions, that w/ a larger number of cards (6+?) dialed to max efficiency, I often had to wait way longer than the normal warm-up window to stabilize on max hash rates - possibly as long as 30min to an hour. I would see the same behavior you are describing, where a couple of the GPUs (different ones every restart) would hash a couple hundred h/s lower for a while, before finally deciding to join the rest of the rig at the max hashrate.
As it would almost always get to the right level sooner or later, and then usually stay there, I wasn't terribly concerned. Though at some point i did start wrapping the miner call in an auto-restart loop w/ a hashrate watchdog, which could sometimes get the rig to the proper h/r sooner.