I played for over three years on chessdotcom and came to the same conclusions as the OP. There is something 'not-quite-right' about the eco system when considering lower-level games - you play higher rated opponents and you see consistency, play lower level opponents and there is no consistency with regard to playing strengths, you could get completely stumped in ways that seldom occurred against higher level players. All this acts as a strong drag on lower-level players ECO ratings such that no progress is made. The main counter argument is 'well that is how the eco system is supposed to work' but eventually I came to the conclusion there was something wrong somewhere.
Never worked out why this would be - although there are plenty of possibilities* - or whether I was deluded, however I no longer play people on chessdotcom. I have not noticed the same 'effect' on LiChess although I play much less on-line nowadays preferring the playing consistency of old dedicated chess computers and Bots.
* My best guess - and only a guess - was that the 300K+ people banned each month on chessdotcom were pretty much the same people month after month, there is no way chessdotcom could detect this for non-subscribed customers. If so, this would leave the lower level gameplay pretty corrupted. Additionally there was nothing to stop stronger players running multiple freebee accounts with much lower eco ratings than the fixed LiChess 1500. Let's face it, everyone likes to win at chess, even when such 'wins' are fairly pointless. The ECO system was never designed to cater for such possibilities or human behaviours.
Note, it's much more difficult for higher rated players to appreciate this, the only time such players would meet sub-1000 mini-Magus players is in tournaments, when the most likely reaction and explanation would be to suspect 'cheating'.
Never worked out why this would be - although there are plenty of possibilities* - or whether I was deluded, however I no longer play people on chessdotcom. I have not noticed the same 'effect' on LiChess although I play much less on-line nowadays preferring the playing consistency of old dedicated chess computers and Bots.
* My best guess - and only a guess - was that the 300K+ people banned each month on chessdotcom were pretty much the same people month after month, there is no way chessdotcom could detect this for non-subscribed customers. If so, this would leave the lower level gameplay pretty corrupted. Additionally there was nothing to stop stronger players running multiple freebee accounts with much lower eco ratings than the fixed LiChess 1500. Let's face it, everyone likes to win at chess, even when such 'wins' are fairly pointless. The ECO system was never designed to cater for such possibilities or human behaviours.
Note, it's much more difficult for higher rated players to appreciate this, the only time such players would meet sub-1000 mini-Magus players is in tournaments, when the most likely reaction and explanation would be to suspect 'cheating'.