There is a more recent Coinmetrics survey on the matter, dating from November 2019 (see https://coinmetrics.substack.com/p/coin-metrics-state-of-the-network-d2e). In this analysis they estimate the amount of lost BTCs to be at least 1,5 million, which differs from the data in the survey referenced in the OP ,and goes to show that survey estimates offer a wide range of values in accordance to the methodology they use.

For example, Coinmetrics details their estimate as follows:

Their approach seems rather much from the technical point of view, and playing it safe by stating those that are "known" to be lost (assuming the Zombie Coins last touched in 2010 do not resurrect at some point), but they really do not go into trying to play around with estimating the number of lost BTCs due to loss of private keys, sending then to the wrong address/chain, etc. Coinmetrics’ survey would therefore stand as a conservative lower bound estimate, probably correct from a technical point of view, but not representative of the real scenario of lost BTCs.