If a merchant processes a fraudulent transaction their processor/merchant bank gets charged for the entire transaction, which they in turn charge the merchant.
More precisely the transaction doesn't have to be fraudulent. The buyer just has to be unhappy enough that they call their bank and issue a charge-back.
So to reduce the chances of that the processors have a risk assessment department that determines the probability of these kinds of transactions occurring. If the probability is high, they just won't work with that merchant.
Now let's say the merchant wants to take money from Party A and give it to Party B. In this case Party B is probably the merchant as they're getting paid for something. But Party B didn't go through the processor's risk assessment process and they may have a much higher probability of causing charge-backs. This basically moves control from the processor/merchant bank to the merchant opening up a possibility for money to siphon out of the processor/bank and potentially never be recovered.
That's why it's very difficult to get approved for something like this. For more info read up on Third Party Payment Aggregation.
I don't think Kerchingle's model is right either. They still require you to click their medallion to begin payment to a publishing site, introducing that same cognitive disruption as conventional tipjars. Also I think the "social cents" thing is basically nuts. Far too easy to be embarrassed. It smacks of that idiotic site that published credit card transactions. Blippy? Blit? Something like that.
The business model I'm following most closely resembles the now-defunct Contenture; Readability is similar but not identical. Obviously I have some additional technology secret sauce or there wouldn't be a project in it to satisfy university requirements.
To me the main failing of conventional tipjars is the requirement to manually tip the recipient. To my eyes Tipjoy was the first real twist on this -- that you could commit to tip before paying -- but otherwise it was the same as Flattr and the dozens of others that have come and gone in this space. Flattr's twist is OK, but I think imposes even more potential cognitive overhead in terms of "I really like this site, I need to find more articles to flattr".
I think that the Contenture / Readability / my-company-name-here model works better because it requires no thought on the part of the customer. Payments are divvied up automatically. No transaction cost ("Do I tip or not?") is imposed on the user.
Mind you, my most imposing business obstacle is getting a merchant account. Businesses which take money from party A and pay some portion to party B are basically viewed as kryptonite crossed with rat poison by the banking sector.
Fair enough, too -- such businesses have statistically high rates of complaints, failures and money laundering shenanigans.