You do realize that 51/41 is 24% more blocks while receiving 370/409 which is %10 less. This is a HUGE disparity that existed even when there were weighting on shares that was proportional at a factor of about 1/8 iirc either way an exponential system can be more easily rescaled easily when compared to the rounding that occurs with fractions I mean how can one reduce 683 * 6/5 + 87* 7/6 + 7* 8/7
You can't is the answer. Rounding is far more harmful for small miners as they are losing out on a larger percentage of their shares over time than larger miners, further more the operational variance is lowered due to the fact that you aren't paying out losers in high amounts. You reward those that succeed not those that want to game the system. Furthermore this system actually works with higher tier chains where as I said before a 16 length chain would only be worth 16/15 which is actually less than 6/5 as your system has the most crucial flaw of not understanding that a a fraction n+1/n will forever shrink down infinitely. Furthermore you seem to misunderstand that a PPLNS has no operational risk which is why ypool uses it not, proportional share values has nothing to do with his. Please do some homework before continuing this line of thought.
Edit: Speaking in terms of optimization, furthermore as mentioned in the system the score accumulated between blocks can be rescaled equally each time a block is found in order to prevent overflow plus it would help to provide a more accurate payout estimate using this system.
Also a note on "fairness"
Let us assume that every single miner were to use the exact same rig and the exact same settings and mine for the exact same amount of time and that the only element of chance was generated through the prime generation. If we used a 1:1 share value in this situation, each person puts in the same amount of effort for finding a block and thus the variance is split between each one and their payouts would be approximately the same regardless of who found the block or who put in what, as long as the number of shares was about the same. Now imagine this community attempting to tackle a length 16 chain. They might spend literally months attempting to get the block and again this would then have everyone payed out equally. Now lets imagine a that in this community someone found out how to submit lots of lower chains so that when that 16 chain block was found they got more than half of the block. Obviously the community now has to weigh their shares to avoid such exploitation, however they only change the weights by a proportional value so each chain down from the target is worth 25% of the previous such that four 6 chains equal one 7 chain. Now this time around that one person does their method of mass chain submitting and while they don't get half of the reward, they still get a larger portion than they should simply due to the fact that it is much easier to submit four smaller chains than 1 larger chain especially over the course of a long period of time.
Now lets look at a more realistic example, if everyone can target their chain lengths and each chain length increase is essentially a ten fold increase in difficulty why would the payout be anything less than a ten fold increase for a higher chain compared to a lower chain? Furthermore this method in fact motivates people to aim for higher chains and further the pool as a whole which mind you, does benefit small miners as they get payed out every block found and thus more blocks means more payouts not just for the giants but also for the common folk. The goal of primecoin pools shouldn't be to allow leaches to steal from those who actually are contributing, if you allow chain spamming then you are essentially enabling pool hopping without even having to leave the pool as it is more profitable to spam low values than to actually aim for higher ones. Furthermore even if xolokram wanted to be nice and later allows a very low minimum chain length for submitting, this system would allow it not to negatively impact anyone and instead would simply allow for a more precise payout/hashrate.
On the topic of min chain length essentially the lower the min chain length the lower the payout variance (if an exponential value system is used) due to the fact that people's work is being accurately estimated however it does raise the operating cost by having to handle an exponential amount of increased submitted shares (regardless of value weighting the amount of 5 chain's found per hour is 10 fold the amount of 6 chains) so with the proper infrastructure in place the min could be set up to 4 levels lower than the target (ie one level lower than it is currently) IF AND ONLY IF the amount that the servers can take before crashing is at least another ten fold of the current. And given the current status of the servers dying pretty much daily I would say that ten times would have to be a minimum.