i myself would probably not use aluminum , because aluminum has a finite work life.
of course it may take years to fail, but aluminum will eventually fail. depending on proper engineering the piece will have a direct impact on how long it will last.
i like the idea of improving mounting of the rocker shaft, it needs to be rock solid, any wiggling means failure sooner than later.
i don't know whether or not i like the roller tip rocker idea, not sure it is really necessary.
Guy is correct in that the tip needs to have a compound curve, to get it to stay close to center of the valve stem. it also needs to be very well polished as it will have a wiping element as well as a rolling element to its action.
in practice tho most folks settle for the center at 50% lift with the rocker traversing from just outside center to just inside center, the goal is to get dead nuts on center thru the stroke.
there are many factors as i have posted before, valve stem height, rocker stand height and position, pushrod length, and cam lobe lift.
it is a series of comprimises to say the least. probably best to actually lay it all out on a drafting table to arrive at the correct dimensions and placement of all components, or to trial fit and cycle thru the stroke taking incremental observations and plotting the results to any single change.
remember change only one thing at a time, and recheck, changeing two or more at once is asking for alot of mind numbing rethinking, rework, replotting and repeated steps with the same repeated results or worse results, and not knowing which change either helped or hindered the effort.
if you are intent upon using roller tips, be sure to use the proper brgs made for cam use, they have thicker outer races, which will be needed because of the valve clearance. the standard thin wall outer races that are meant to be supported will quickly fail (shatter) and cut up your valve caps.
i think before i would go to all the work designing a new valve rocker train i would learn from the original lister rocker end profile and try to duplicate what they engineered.
likely it could be done and the result would probably outlive anythng most of us are going to accomplish trying to reinvent the wheel.
bob g