Maybe Tiburon doesn't see it as a problem?
Now I can't speak so much for CFM and Play Now, but the AI system is capable of being used to create a desired effect.. This is very prevalent in MUT Challenges; it is a staple of the MUT system to have the AI play a particular way to force the user to win by design, and to create challenge and that tense game experience..
For example - if the challenge involves the user to score in under 2:00 minutes, the start score for user will typically be 6 points below the CPU score, so a TD is required and the extra point leaves the CPU with just a FG needed to win...IF the user doesn't use all the clock, the CPU comeback ability is absurdly good; typically if the CPU has more than ~40 seconds left, you can expect a FG and a loss...
Now, why would this type of AI scripting be necessary? Well, follow the money...MUT is about making more money...That money is made through selling card packs...How can a company sell more product? By making the consumer believe they need it...So, you lose challenges in MUT and the user can be led to believe they just need better players.. Well, you could grind out challenges to earn coins and win player card packages, but you have to win first, so still the user is to believe his team needs to be better "now" to win, so buy player packs... The fact that money is involved is enough to be suspect in regards to rubber banding or comeback engine play... Just follow the money trail...
I have a 92 OVR MUT team, and in a challenge currently available in the Limited Time challenges, against the Raiders, they have an 83 OVR team with a skill level of AP...I can't score too early, and many times I can't make the win. If I score too early, you get a lot of what I mentioned above; a great drive to a FG..Now, I have beaten the challenge, but I have also replayed it multiple times for practice and just to see what I can and can't do against the AI, and I am more likely to lose than win on a per game basis..
For example, the MUT Raiders Corner Sean Smith (85 ovr), 84 speed, 81 M2M - going against my 96 OVR WR Emanuel Sanders (w/chems active) 93 speed, 96 route running, as an example, cannot beat Sean Smith in M2M cover on streak and fade routes.. The CPU will not allow this to happen - UNLESS, I manipulate the AI with route adjustments at the LOS to confuse the CPU....So the game teaches users to manipulate the AI to win..
Now, on Rookie and Pro, the above example matchup would result in a win for Sanders, all day, and it happens..From AP and AM though, the AI is set to not only be more difficult obviously, but to manipulate events in game to achieve any challenge desired..And in MUT you usually have to face AP and AM levels to get the big player cards and rewards.. I assume that such manipulation is not as prevalent in CFM and Play Now because the game situations are broader and thus harder to control, but in a set MUT Challenege, where the game specifics and goals are already outlined for the user, the AI can be made to present any effect desired.
Note - This is merely my observations and opinions, nothing more..I have no concrete proof except for what I have presented above; theory..
ANother issue to consider - Ratings, not that diverse..THus the difference between 84 speed and 93 speed, for example, isn't much of a gap; but the numbers leave people "believing" there is a gap, and thus also would create the mentality that they need to purchase better players, even with slightly better numbers..