You're measuring the cost to the government, not the cost to the public. How much time do people waste filling out eligibility forms? What's the economic cost of wastefully converting fungible money into non-fungible requirements to buy specific goods and services?
Meanwhile the "extra spending" isn't real because the benefits, being universal, are going to many of the same people paying for them, and therefore simply cancel out. The purpose of making them universal is the simplification: It's simpler -- and therefore harder to cheat -- to have a UBI and a uniform tax rate than to have a benefits phase out that effectively cancels out the lower rates on lower income people in the existing tax code. Or worse, creates marginal rate cliffs and highly irregular rates because overlapping benefits phase outs interact with each other in random and perverse ways.
You are quite correct that the overhead of most government programs is relatively very small.
But look at your numbers....12% of the population on SNAP, 20% of the population on Medicaid, etc etc.
What would it really cost us to just pay the equivalent cash to everybody, and then raise the taxes to offset the costs on people who don't need it?
We are already paying for--piecemeal and very inconveniently to be sure--much of what we'd want a UBI to give us. It probably would cost us more in taxes than what we are paying for now, but it wouldn't be an apocalyptic increase. The delta of spending is not as big as what one would initially think.
Administrative overhead is nowhere near as high as you estimate.
SNAP covers ~12% of the population with 5% overhead costs.
Medicaid covers ~20% of the population with 5-10% overhead.
Social security covers ~20% of the population with <1% overhead.
The extra spending to make benefits universal is an order of magnitude higher than any savings from eliminating bureaucracy.