In foobar that's how we'd do the string simplification for common tag fields - easy (untested code, just to give an idea) :
$replace($lower($ascii(%album artist%)), ,,&,,",,'',,(,,-,,_,,),,=,,#,,{,,[,,\,,@,,],,°,,+,,},,$,,£,,%,,*,,',',,;,,:,,!,,?,,.,,/,,§,,)
$replace($lower($ascii(%album%)), ,,&,,",,'',,(,,-,,_,,),,=,,#,,{,,[,,\,,@,,],,°,,+,,},,$,,£,,%,,*,,',',,;,,:,,!,,?,,.,,/,,§,,)
$replace($lower($ascii(%title%)), ,,&,,",,'',,(,,-,,_,,),,=,,#,,{,,[,,\,,@,,],,°,,+,,},,$,,£,,%,,*,,',',,;,,:,,!,,?,,.,,/,,§,,)
Hope you get the idea. If we can do the same for the lyrics web sites data and compare the resulting strings, we'll have solved the problem, don't you think ?
BTW it would be nice if Lyrics Grabber could parse the %album artist% field BEFORE the %artist% field. I have the feeling that it only parses the %artist% field, which can lead to errors in many songs (e.g. all duets where %album artist% = name1 and %artist% = name1; name2)