Upon further reflection, I would think that using the current top bid or ask would be far better than last sell for thinly traded coins, as the last sell will ping-pong back and forth between the buy and ask prices which may have considerable distance from each other. This price jumping causes a significant oscillation in your reported in-process figures from report to report that obfuscates the underlying mining profitability results (in the short term). I would probably select the highest bid, as you are naturally a seller so this is the true current value of your holdings. A miner that truly wants visibility to his mining proceeds at a finer level considers in-process mining balances in addition to their paid BTC, which yields a more accurate conclusion about the profitability of your pool when making short term decisions (like whether to stay after a day or 2, or how am I doing today). When you factor in the in-process proceeds you naturally get a more stable profit reporting than payouts or balance alone, but not if it's too jumpy.

For example (just to be clear),
24 HOUR PROFIT = Today's paid BTC + SUM(current in-process balances ) - SUM(in-process balances 24 hours ago)

A small point, and just my 2 cents.

On a separate note, it would appear that you have now implemented PPLNS type logic or at least logic that has a similar effect. As an experiment, I had my son point one of his rigs at your pool for a 48 hour period to a new wallet address. At the same time, he already had an identical rig mining on your pool for many prior days for which the returns had already stabilized.  When the new rig was added, hourly productivity was compared between the two identical miners. The second (new) rig's return was very low for many hours as compared to the existing rig (factoring in all balances and mining side by side), and then gradually came up to a similar level of hourly productivity of the first rig after a day or so. To complete the experiment, the new rig was then removed, and its wallet continued to earn new mining proceeds (immature, etc) from your pool many hours after it was no longer mining.  Even with the after-effect, total returns from the second rig never quite reached (in aggregate) what the first rig earned during the 48 hour mining period. This is only a single data point with other unmentioned minor factors and as such is not statistically valid, but have you implemented something that would have this delay effect? If PPLNS, how do you calculate since there are many different coins and productivity is different for different coins for the same miner?