Lowering Queue times is not an end in itself, it's a means to an end, and it fails

Firstly, the ultimate end goal of Riot Games is to make money. They are a company, after all, so by definition the goal is to make money. One aspect of the money making is to get people playing the game. Lowering queue times is a means to that end, intended to make people play more games, thus running into and gaining an interest in more champions, leading to them buying more champions and skins.

However, their methods of lowering queue times are at odds with the higher-level goal of making people play the game because they contradict the more impactful goal of making the game enjoyable:

• They've decided to match people with a broader pool of players, further from their own skill level in both directions (higher and lower)
• Autofill-- which most often means Support, but can vary patch to patch (such as when a new support is introduced to the game)

Matching people with players with lower skill than their own, while the enemies are higher, leads to matches where they're frustrated at allies who are reasonably struggling against their lane opponents. And frustrated at themselves for reasonably struggling against their own lane opponents. In general, one or several such frustrating games is likely to make someone not want to queue up for another match, thus defeating the purpose of lowering the queue time. This also has the unintended effect of skewing the data about queue times, since a player who logs off is just as not-in-game as someone waiting an extra three minutes in queue.

Autofill lowers player enjoyment for several reasons. The most obvious of which being that you clearly don't want to play Support/Jungle/Bot (in that order, generally speaking based on popularity of the roles) if you queued up Mid/Top. Then if a player dodges, they're only increasing the likelihood that someone else gets autofilled and dodges, usually repeating several times until someone decides they'll take one for the team... but with the way having one player who doesn't know their role works, it's actually taking one against the team, causing both them and their allies to have a bad time. This secondary effect of the players allies having to suffer through having a support/jungle/bot who isn't up to par would be another reason autofill lowers enjoyment for players.

When these two things, and just as likely both at the same time, are at play, people end up having games that aren't fun whether they're a win or a loss. And the frustration comes not from a "bad" game but from having a game where you had no control over the factors that made it bad. Then, after such a frustrating game, you're less likely to actually queue up for another. This leads to a lower number of online players, which simultaneously exacerbates both problems and makes it look like they need to be applied more, because they're supposed to be acting as the solutions to a decrease in online players.

Now, as far as solutions go it's first important to note that it's irrelevant to bring up the fact that blind pick was once the only method of matchmaking. That caused the exact same problem that Autofill now does; it forced people who got last pick into whatever role no one else wanted. The solutions are rather clear; constrict matchmaking criteria and remove autofill. Yes, this will lead to an increase in queue times, but it will also lead to an increase in player enjoyment. I'm sure most players would rather wait a little longer to be guaranteed the role they actually want to play, and opponents/allies closer to their relative skill level, than to be forced to play support against a diamond 1 support main. And it's this increase in enjoyment that will fix the issue of fewer players logging on, not forcing players to play unfair matches in roles they don't want.