Few things missing.
System weight, as a percentage, is a significantly larger tax on lighter riders.
Rolling resistance is a greater percentage of loss for lighter riders.
Mechanical loss is also a greater percentage of loss for lighter riders.
There’s been quite a few studies on this. A few magazines and YouTube channels have attempted to compare the two. Usually, the results surprise the testers.
At equal w/kg. Two riders, one 60kg, one 80kg. The 80kg rider will have an advantage on all terrain. Everything. Remember, equal strength riders, by w/kg.
Flat, larger riders have more muscle mass, it’s not a penalty, even in regards to CDA. I’ve read many attempts to quantify it. It’s incredibly variable due to a million reasons. However, hypothetically two identical riders, scaled perfectly up. The larger rider usually has a .25-.50 w/kg advantage on a flat road. Muscle mass vs CDA. It’s an equation, something to do with cubic volume vs frontal area. I’ll attempt to find it. It’s real.
Cannondale did the climbing, gravity vs aero science in their Super Six white paper. The crossover to advantage gravity vs aero is related to ultimate power. For a 3w/kg rider, it’s a low gradient. Something like 5%. For a World Tour athlete, say 6w/kg, the crossover is far higher 8/9/10%. They are travelling so much faster that aero remains a large factor right up to high gradients.
Take a real world example.
60kg 4w/kg rider
80kg 3.7w/kg rider
The lighter rider is slower nearly everywhere, flat, downhill, false flat…
In order for the lighter rider to be faster the gradient would need to be nearly all of the power requirements. So likely 8% or above. The lighter rider has to overcome the bike and system weight being a greater percentage of their power, rolling resistance and mechanical drivetrain loss. At .3w/kg it would be very close to a wash.
Now, a poster above made a fantastic point. One many overlook.
Fueling.
The longer the event, the greater the advantage for the smaller rider. On the proviso, it’s a road race and they are able to draft on flats and downhills etc.
Imagine 2 riders, hypothetical situation, same fat oxidation ratios, same aero shape, same humans, one is just proportionally scaled up to 20kgs heavier.
Both riders hitting the magical 100g/h fueling.
6 hours into the race and the 60kg rider has taken in 600g of carbs. 2400 cal. They’ve used say 3600 cal to get to the final climb. They have 800 cal of glycogen left for the final climb effort.
80kg rider has used say 4800 cal to get to the final climb. Same 100g/h fueling. They have 200 cal of gylocogen left for the final climb.
They are in a greater deficit.
Now, the larger rider can intake more cal, they can store more glycogen, however it is not anywhere near as high as it should be, relative to the smaller rider. Whomever built us got the math wrong 
Compound this with the fact that many heavier riders don’t exceed 100g/h. If they were able to tolerate and absorb, say 130g/h it would be awfully close.
Takeaway. Larger riders need be certain they are at max fueling ratio in races. Particularly, in longer events.
In regards to climbing. I’d say for most competitive amateurs the crossover is the 8% realm. That’s just a seat of the pants measurement for 4 to 5w/kg riders. That’s were aero becomes negligible. Gravity becomes the vastly greater percentage.
Personally, as a light rider 60-62kgs. I would never attack etc on anything less than 10%, if I was in control of the racing situation. Even then, the race would have to finish on the climb. As if it didn’t, I’d likely be caught on the downhill or run in to the finish.
The way many people ride, particularly group rides where hills are raced gives many the feeling that w/kg is everything. Obviously, it’s very important.
I don’t think there’s an exact grade where absolute power trumps w/kg. As the gradient angle varies in a ratio to speed of the rider.
I’d be comfortable with 8%, as a estimate for most recreational riders. Anything less, aero is a significant factor, so raw power is still an advantage. If there’s drafting in the equation, I’d say 10%.