Both Buy and Sell SAFE and Minimum Share Transaction values are based on the current Share Value in my spreadsheets, in keeping with Lichello's AIM model. I ran some comparisons on historical data using Tom's (OldAIMGuy) suggestion of using these values based on Portfolio Control in lieu of Share Value, and found no change to the overall returns. It made the transaction size more consistent whether price was rising or falling, and while that seems like it would produce better results, it doesn't. It benefits on one side of a trade and detracts on the other side (as compared to Lichello's use of current share value), so the results effectively offset each other.
Lichello used the same SAFE factor for both buys and sells in his original model. Split-SAFE allows us to tweak the algorithm a bit to try to improve results of the original model in certain market conditions. One example Tom suggested that has significantly improved returns in my backtesting is increasing Buy SAFE by 5% for each purchase, and resetting to the original value on the first sell. This is used only on investments such as some 3x leveraged ETFs whose historical chart shows large (>60%) declines in share prices during market downturns, where maintaining the initial Buy SAFE value will result in cash depletion. The adjustments allow the cash to be conserved a bit more and allow lower-priced buys, and it significantly helps Black Swan trading if cash is depleted near the bottom of the trend. Running out of cash at or near the bottom is the ideal AIM scenario and will result in maximum gains when the price recovers.
In most of my testing, the 5% Minimum Transaction Size improves annual returns about 0.5-1.5% but almost doubles the number of transactions. In a few tests, the 10% setting resulted in slightly higher returns. In my testing on an investment that had Buy/Sell/Min set at 10/10/10 and it traded sideways and had no trades for several months, I'd temporarily change to 5/5/5 to generate a few trades and return to 10/10/10 when prices started moving away from the narrow band. It's better to have a few trades in these circumstances than months of no activity. It's a similar strategy to the Black Swan trading during periods of cash depletion. Every little bit of AIM trading helps the overall return.
I'll run a few backtests on what I consider the poster child of a mostly upward trajectory ETF, the TQQQ - 3x-leveraged Nasdaq-100 Index. This one is hard for AIM to beat the buy-and-hold strategy, or if AIM wins it's not by a lot. I'll run the standard 10/10/10 scenario over 5 years and a couple where SAFE values are adjusted. I'll post the results when I'm done. Hopefully that will help answer your question and offer the board some additional insight.
Tom and others have offered their experiences and how they use the SAFE factors. As has already been discussed, often the strategy used depends on one's goals and the investments selected. Highly volatile investments like 3x-leveraged ETFs could employ different strategies than a non-leveraged equity or ETF investment. Some might want to significantly raise the Sell SAFE factor to restrict selling in a rising stare price when cash is excessive, where others may prefer to spin off some of that excess cash and share value into another AIM-managed investment.
One thing to remember when adjusting SAFE or MIN factors in my spreadsheets.... it only impacts future transactions. It will cause all prior transaction calculations to look strange, but since the 4 blue columns for manual entry documenting each completed transaction is unchanged, it has no impact on AIM function and overall results. I expect the same is true in other AIM spreadsheets.
I'll report back when I finish backtesting TQQQ. Or is there a specific investment you would like to see tested that is better suited to your questions?
Market Data 
Markets 
