Mean Variance in Practice - Questions from UG student

Dear monkeys,

I am doing an investment and trading simulation for a class in my undergraduate degree, and have a few questions regarding the mean-variance approach to portfolio construction, which I hope you practitioners will be able to answer.

The universe of investible assets within the simulation are equities, bonds and cash. Within equities, there are different sectors (oil and gas, consumer, pharma, tech with respective indices and underlying stocks...) and within bonds we have 2y, 5y, 10y and 30y government bonds. I would like to set the SAA of my portfolio with a mean-variance approach and then tactically overweight the sectors I like the most.

My questions are:

1. To plot the efficient frontier, can I use the mean return of each asset class (i.e. the average of historic returns)? Or must I use the CAPM?

2. Does it make sense to use the below as asset classes?

3. Using historic returns looks sensible for equities as we have over 10 points of data (1 per year). However, we only have 3 years of data for bonds. Because the realised returns of bonds have been negative in this period on average, the mean return for these is negative. I feel like this doesn't make sense. What could I use instead?

4. Once the SAA has been determined, can I only overweight or underweight asset classes by buying more of the index or can I select stocks that I think will do particularly well?

Thanks a lot for your help!