Focusing on one project should not actually be more beneficial?

People often say that focusing on one (CPU or GPU) project is most profitable. According to how I understand the calculation of MAG, this should not be the case.

The magnitude for a project is calculated as

MAG = ((Researcher_RAC/Team_RAC)/Number_Whitelist_Projects) * 115000

Number of whitelisted projects and the multiplier 115 000 can be defined to be constants, so the only deciding factor is the proportion of your RAC to the Team RAC. Of course one needs to notice here, that Team RAC also includes your own RAC. So we have

Your_RAC/(RAC of others in team + Your RAC)

For easier denomination, lets use

W = your own RAC
A = RAC of others in the team.

Ultimately if your own RAC is very small compared to the team RAC (W<<A), you can approximate

A+W ≈ A

So your mag is proportional to

W/(A+W) = W/A.

This mean that if you multiply your RAC by 8 (e.g. number of cores in your CPU), your MAG is also multiplied by 8. Easy peasy, right? In this situation the rewards grow linearly, so it doesnt really matter if you focus on the project or not.
But lets look at the situation when you are a large contributor for the team in the project. Lets assume that your one core produces one tenth of the RAC of others in the team, i.e. W=0.1A
This leads to your mag to be proportional to

0.1/(1+0.1)=0.0909… ≈ 0.091 = 9.1%

So what happens if you use all 8 cores on one project?

0.8/(1+0.8)= 0.444… ≈ 0.444 = 44.4%

If we divide the bigger contribution by the smaller one, we get

0.4444/0.090909 = 4.888…≈ 4.9 = 490%

So, by using 8 times the cores, we see only a 4.9 times increase in magnitude. This is because the mag for a project is capped, you are essentially competing for magnitude against yourself. Of course if an other projects has much higher Team_RAC, it might not be worthwhile to change.

Mathematically, focusing on one project should not be benefical.

So what gives? Why do people say that you should focus on one project for max benefit? Honestly, I don’t know the exact answer. Either people just feel that they are getting more by focusing (bigger single mag number to look at and confused by the effects varying RAC of yourself and team), or there is something in RAC or decay calculations that gives an exponential growth to your RAC based on credits done.

If someone can point me where I've gone wrong, please tell me in comments or in a PM. I've been confused about this for a while and would appreciate some clarification.

If on the other hand my calculations are correct, people should be able to mix and match projects that are close in Mag/work, assuming your work units dont get old because of it. You should still keep CPU and GPU projects separate, to not waste your CPU computing power. Bigger players with huge mining rigs could at times benefit a lot from mix & matching. Of course the amount of GRC per day is constant, so all these converge to the same total in the end. But I think many people would like the freedom to contribute to and try out multiple projects.