If we subtract Satoshi’s 1 million BTC (as Plan B does in his STF2X model), should we also subtract the ~4 million BTC estimated to have been lost?
If you subtract Satoshi's bitcoins, you should also substract the ones that are same like inactive, correct. The questions is: Should you subtract Satoshi's bitcoins? Should you trust a study that showed an estimated number of lost bitcoins? Should you trust Satoshi for not being alive and not being willing to spend them? We've got rid of trust in this blockchain-lized era, if something doesn't have a proof that it's gone for good, it is simply not gone.
For example, an OP_RETURN output is a solid proof that these coins are burnt; a burning address is also a proof that no one could generate such RIPEMD-160 hash, so no one owns it and therefore, they're also burnt.
Although, I had also read that around 3 million BTC, excluding Satoshi's, have left the circulation. Make your own research and your own assumptions.