Indeed so the methods could be precisely accurate or imprecisely accurate or precisely inaccurate or imprecisely inaccurate. Hence as BFTV mentions above consistency of methodology is key.
Once there is enough data from new methodologies, comparison, correlation and extrapolation can be made, but with the caveats re precision and accuracy being important.
It's a bold move to bin an entire dataset due to a new way of testing / evaluating. If we look at sun spot No's, we know there were inaccuracies in the data, but the "new improved" data hasn't just trashed the data from e.g. the maunder minimum, because there has (I believe) been a correlation exercise which takes the limited equipment etc into account. Maybe something similar could be achieved with ice extent, thickness and age by running methods alongside for sufficiently long to be meaningful.