Starting situation
Suppose a name and birth year are known, and several records share the name. Compare the confirmed year where present. Do not use a returned age as evidence before examining its time context.
Another signal
If an independent source supplies another identifier, check whether it matches. Do not take a random contact from one candidate and promote it to an initial fact. That would merely reinforce your own assumption.
A sound outcome
If no record has enough support, the outcome remains uncertain. That is not a methodological failure. Record what matched and what is missing, while keeping candidates separate.