Conditionnaly remove duplicates

Split the table into WEB API and not WEB API rows. In the former, remove rows with IDs that exist in the latter and then append them together.

Try pasting this into the Advanced Editor of a new blank query to see the steps:

let
    Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("i45WMlTSUQrwcfQLUXAEsnwjgwN9lGJ1opWM4OJOhrgkjFAkjOESzijiJnBxFyAr3NVJwTHAE90oYzQpU7iUK7JMLAA=", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [ID = _t, NAME = _t, SOURCE = _t]),
    #"Changed Type" = Table.TransformColumnTypes(Source,{{"ID", Int64.Type}, {"NAME", type text}, {"SOURCE", type text}}),
    #"Filtered NonWEB" = Table.SelectRows(#"Changed Type", each ([SOURCE] <> "WEB API")),
    #"Filtered WEBAPI" = Table.SelectRows(#"Changed Type", each ([SOURCE] = "WEB API")),
    #"Removed Duplicates" = Table.SelectRows(#"Filtered WEBAPI", each not List.Contains(#"Filtered NonWEB"[ID], [ID])),
    #"Appended Query" = Table.Combine({#"Filtered NonWEB", #"Removed Duplicates"})
in
    #"Appended Query"

For large datasets, it's probably more efficient to use an anti-join rather than List.Contains but the idea is the same.