Of course you can put as many free tx in a block as you want. 1 more, 100 more, 1,000 more up to the 1MB limit. However the larger the block, the longer the propogation delay. The longer the delay the higher the orphan rate.
Once again miners certainly "can" include more free tx but doing so would result in less net revenue. How much less well only a large mining pool experimenting (in a very cautious way) by comparing rate of orphans to size of block would know for sure. Others could analyze from the blockchain but it is harder to get a clean comparison.
Still miners have no real incentive to try an push the orphan limit for free tx. 1) it just encourage less paid tx and 2) it increases the risk of orphaned blocks (which is worth 0 BTC) how much more is debatable* but it most certainly is more.
* meaning is it that a 500KB block vs 250KB block results in a 0.1% increased orphan rate or 1%. That might not be known but if it is >0% then adding 0 additional BTC simply results in less revenue. (25+x)*y% < 25 where x=0 and y <100%. Now if x is >0 then there is an equilibrium point but where the increased size warrants the increased risk but if x =0 you are just risking more orphaned blocks (loss revenue and higher volatility) for literally nothing.
And obviously a mining pool should not risk to do it alone, if it's to become a reality, there should at least be some network-wide consensus, so everyone takes the same risk and it evens out.
That will never happen and likely shouldn't. However as the subsidy continues to decline and tx volume rises fees will become more and more important. Many pools simply keep all the fees now so members of the pool don't really care one way or another. However after the next subsidy cut when a block is worth 12.5 and if tx fee volume quadruples to 2 BTC per block fees starting becoming a sizable portion of a miner's revenue. I would expect mining pools to provide data like average fee per block and miners will gravitate to various pools which meet their risk/reward ratio. One pool may just scoop the largest fees and keep blocks small. Sacrificing some revnue for reduced volatility. Another pool may scoop every tx with a fee greater than 100 satoshis to pull the maximum revenue (and maximum volatility).
As fees, block size, and orphans become more important I imagine the number of free tx accepted will likely decline (or will only occur during periods of low tx volume). The fact that miners accept ANY significant number of free txs is merely a distortion of free market caused by the subsidy. As the subsidy declines the effect of that distortion will decline as well.