The raw facts I've just outlined don't disprove your thesis, but as I understand it the details do. It's also the fact that outside of situations where politics are in play, expanding bullets are the universal choice, all things considered.
Note this includes civilians who carry, there's 7.5 to 8 million as independently estimated last summer by scholar Clayton Cramer and the GAO respectively, plus those who can carry concealed without a license in the handful of states that allow that (Vermont, Alaska, Arizona and Wyoming). Unlike the police, we don't have legal or political protection from accidents, we own every bullet we fire, and a loved one could easily be one of those too close bystanders (or on the other side of a wall or the house, whatever).
Maybe we're wrong, but using hollow point ammo is what almost all of us have decided to do.
I figure innocent bystanders who get shot are hit by fewer bullets than the people who are actually being aimed at. Pulling numbers out of my ass, say innocent bystanders that are hit are hit by 1 bullet on average, and crims that are hit are hit by 3 bullets on average.
The questions to me then become:
1) how many innocent bystanders are being killed by those single bullets, and how would that number change if they were FMJ instead?
2) Given the average number of times crims who are shot at are hit, how likely is it that they would still be killed if the police were using FMJ?
I suspect these two factors work out such that FMJ would be slightly more dangerous to police, but vastly safer for the general public despite the increased penetration power. Police are signing up for it, so I am okay with putting them in more danger to keep the public safer; I am sure police unions would throw a fit over that though.
(I think it would also be revealing to examine if perhaps the lower penetration power of fragmenting rounds make police have a more relaxed attitude towards accuracy.)