There is not a schread of evidence, anywhere, that there is any more than 2% power increase on an apples meat apples basis.
The info I have from Yella Terra, Crane, BRC, SVO, none of it defines the gain scientifically.
The are real subsidary benefits attributable to the being able to optimize the combination because of roller rockers (side clearance reduction, blowby reduction, no galling on the exhast valve when you have roller rockers, the ability to run practically no valve guide into the intake becasue of no side loadings, being able to get a season out of a 320 Isky cam in a Pinto 2000, rather than six races, yadda yadda yadda).
The three engine I am familar with that benefit most from roller rockers are A-series Minis, Ford SOHC Pinto and SOHC Limas, and the heaviest valve gear engine ever, the 351M, 400 and 351C '335" series engines. All have terriable durability problems with big cams when that have no roller rockers. There is a potential for a 25% boost when roller rockers are used and the package is built around them, but they are a catalyst, not a source of power.
There are funadamentally five gains I have read about.
1)The difference in sled fulcrum frictions verses needle bearing friction and (specific to some ohv pushrod engines)
2) reduction in side loading friction in each valve.
3)The side wipe from the machined face of the rocker against the OHC cam lobe on some Lima/Pinto/Romeo engines and,
4) on all engines, the almost total reduction in side wipe against the valve tip.
5) The reduction in required valve seat pressure (averting spot weld valve seat loads and overloading the pushrods and reduction of critical loads on all parts of the system). This is a huge gain in long distance endurance events. The first use of roller rockers (illegally) at Bathurst by Alan Moffat in the seventies capitalised on this facet.
These five items are the only both only machanical benefits. All friction is manifested as heat out, so the gains are related only to these to factors.
No disrespect to any maker of roller rockers, but there benifits are very small 2 percenters on street engines at the time of dyno testing. At 50 000 miles, I sure there could be a 10% difference. A roller rocker is a 'catalyst' or 'portal' to optimizing the rest of the engine, not an absolute source of power in themselves.
I'd never used them on an engine unless there engine was built around them to capitalise on there strenghths.