Alfred has a fatal flaw.
Alfred indexes items from a variety of sources: applications, messages, files, scripts, etc. When a user submits a query, it uses an adaptive algorithm to rank the items in terms of the likelihood that the user wanted this item.
However, Alfred's matching algorithm is awful. If you execute the same query repeatedly in series, the adaptive algorithm reaches different conclusions about which items you wanted. Even if you TELL it: "When I enter the query 'f,' I mean that want THIS SPECIFIC ITEM," Alfred will ignore your instruction and give you other results that happen to include the term "f". The item you specifically requested *might* be the first result - or it might be the second result, or the *last* result, or nowhere in the list.
This bug has been brought to the developers' attention. They are enamored with their adaptive algorithm, so they do not care and will not fix it.
Don't use this garbage software. Use another package that respects users' instructions more than its own haphazard guesses.