Overhead isn't applied uniformly. For example, tuition for Ph.D. students isn't charged overhead, nor is (usually) equipment. So on $1m of funding, if you've got 4 Ph.D. students, that may be something like $200k/year of tuition that isn't subject to overhead. Add in another $100k of equipment and suddenly that 50% indirect cost rate is actually more like 35%, so you end up doing $1m of "work" on $1.35m of budget.

Departments often negotiate something called "overhead return", which is a way of returning a small amount of money to the individual departments -- some of this does things like supporting Ph.D. students if their advisor runs out of funds, or helping research faculty bridge short funding gaps. These things are reasonable and help the institution remain coherent through the uncertainty of grant-driven existence.

There's waste everywhere, but it's not quite as bad as it might seem without a deeper understanding of the university research funding model.