I don't know whether it has always been like this or whether this is new behaviour, but searching for names that have been transcribed using UCF characters doesn't always work. Here are some examples:

Search for burial of Sarah MARSH in Burslem, Staffordshire in 1730
Surname transcribed as MAR*
search_record

Search for burial of Thomas FREE in Barlaston, Staffordshire in 1684
Surname transcribed as _REE
search_record

Search for burial of Mary COOPER in Burslem, Staffordshire in 1720
Surname transcribed as C__PER
search_record

The above all happen to be burials, but that's not the problem because I can find this burial:

Search for burial of Thomas TITLOKE in Cannock, Staffordshire in 1671
Surname transcribed as T_TLOKE
search_record

But here is another nasty problem: we are not being careful with word boundaries when matching a UCF string. If a surname is entered as DO_E, that string of 4 chars should be treated as if it had a word boundary at either end, meaning that DOVE and DORE should match, but not ANDOVER. Here is an example of a wrong match:

Search for baptism of Anne ANDOVER in Cannock, Staffordshire in 1667
Surname transcribed as DO_E
search_record