Half the information on CDs is analogue

The pits only serve to provide digital iformation which requires translation. The reason for integral or external DACs is to translate the digital (not analog) to analog otherwise you couldn't get anything from the pits. If you mean analog is defined as something physical with a surface that provides information in a digital format. Then OK I guess? but that is not what analog means to virtually everyone else.

One of things that I was refering to is the "analog eye pattern".
Contrary to popular opinion, tranports take analog information from a CD and 'build' a digital signal. Only then is the digital signal sent to the DAC for conversion back to analog.
The sound coming from our speakers is only going be as good as:
1. the transport's ability to pick up the raw analog input signal as a reflection of the laser bouncing off the CD (the 'eye pattern').
2. the transport's ability to use this analog information to create and generate a digital bit stream that is a close approximation of the digital bit stream that existed in the recording studio.

The CD data retrieval process is actually very different from phono playback.

Under normal conditions the CD transport will retrieve an exact copy of the digital data used to create it (think CDROMs - computers won't tolerate a "close approximation" of the original data). The CD contains a large amount of redundant data in the form of Error Correcting Codes (ECC).

Data is not read from the CD in a linear fashion like a record groove. The data is stored along and read in in blocks. Each block is protected by ECC code and contains subcode (including timecode - this displys in the transport time window).

Also the laser doesn't translate individual pits into bits. Instead pit patterns, called symbols, translate into bit patterns. These pit patterns are designed to be easy for the photodector to read. Finally the blocks of data are not recorded linearly but are reordered and spacially seperated from each other on the disk. This reduces the likelyhood of a scratch rendering the disk unusable.

The transport reassembles the data blocks in the correct order and uses the ECC codes with a mathematical algorithm to detect and correct all single symbol read errors and detect almost all multiple symbol read errors.

For multiple symbol errors the player may resort to error concealment, where it generates some best fit data to fill the gap. Worst case it mutes the output momentarily (a skip). Multi-symbol read errors are extremely rare on disks in reasonable condition.

The biggest issue with transports seems to be clocking accuracy of the output datastream. With proper engineering this can be almost independent of the physical reading of the disk. The transport can read ahead of the actual listening point, buffer the data in memory, and clock it out at a highly accurate rate.

Good eye pattern helps the accuracy of symbol reading and reduces the need for error correction.

Because the disc is, in a sense, an analog representation of a digital data stream, the raw data is imperfect. But we don't listen to raw data. Corrupted data is to be expected, and the signal information is encoded with an error correction algorithm that yields "clean" data.