Straining out a gnat & swallowing a camel here.
There being some adjustment to consumer habits over time is a given. People buying fewer pagers than they did 25 years ago is still vastly different than saying prices haven't gone up because peoples' shopping carts full of rice and beans during a depression (derived weight given consumption) costs about the same as a cart full of steak, fresh produce, & other perishables that were the grocery basket of goods from a booming economy a few years prior.
And that's WITHOUT the dozen or so manual "adjustments" that can be made outside of any formula. Jim at BLS calls Rhode Island grocery store to find out what their price is on a dozen eggs, and nobody answers? BLS assumes/estimates. Store changed inventory/type/quantity since the last input? BLS manually chooses whether to accept, exclude, or modify. Outlier found in data that's not within an expected range? BLS can determine it was a fluke & exclude it. Data from this month produces a crazy spike in some area? BLS can tweak formula.
There are obviously difficulties with a pure numerical expression of exactly how much more of a fiat currency it takes to trade for a changing array of goods & services. But like all of them, CPI is absolutely filled with guesswork, assumptions, human bias, and a plethora of ways for a government with a financial incentive to keep the numbers in a certain range to nudge the outcome in a certain direction.