Real World Answer:
The discrepancies are a series of continuity errors.
Several issues here.
- Relay stations have limited bandwidth.
- Subspace communications have limited range.
- Subspace communications have a known finite speed, TNG WF 9.9997
- Not all communications are the same distance.
- There are multiple bands.
- Weak signal strategems seem to work with subspace communications fairly normally.
We know about subspace communications relays because of a rather fun TNG episode (Aquiel). We see a relay station. Its role is discussed in episode, as well. Suffice it to say, it's a standard relay technique. Routing through a relay net takes some time, but often allows faster communications once the routing is established. As an example, it often takes several seconds to establish a route on the internet, but once established, the individual packets take a few dozens of milliseconds to travel that route. While the relay has a huge bandwith, it's presumable that those relays also have massive amounts of use.
We know that the individual communications have limited ranges because of both the inverse-square law, and the existence of subspace relay stations. We know the inverse square law applies because it is inherent to the nature of a wave in a 3-dimensional medium.
We know subspace communications have a finite speed because the episodes imply so, and the TNG Tech Manual says so (p. 111). There are times when it's taken significant lag to cover a certain distance in both directions. We also know that it travels in a wave.
We also see many of the communications with higher command being to local bases, not all the way back to SFHQ. Communications with the local base almost never show lag.
The various bands of subspace have different propagation speeds; Paris mentions this in at least one episode of Voyager. If maximum frequency correlates also to propagation speed, then the various bands also have different data rates.
We see and hear various radio-centric strategies for bosting long range communications, and they tend to match real-world radio modes... boosting signal power, moving out of the line of interference, multiple antenna interferometry, signal amplifiers, digital reconstruction. It's safe also to presume that frequency division of a digital signal can be used to make a borderline signal readable at range by sample averaging, albeit sacrificing bandwith.
It's safe to presume that this all adds up to being able to, in certain conditions, get a high-speed comm link, while in others, having to use slower and more relay-based bands for slower comm. We can also make a supposition that the relays pretty much always use the fastest bands from relay to relay.
It's not a large stretch to assume that the faster bands also require more energy; this would mean that the best communications are going to be, in most ship's cases, limited mostly by the ship and its nearest subspace relay.
Also note that communications lags really don't get noticed until they exceed 0.5 sec, since it's pretty typical for people to take over a second to begin to respond, and even a 1 sec delay isn't one that many people would notice in a conversation they are merely observing.
FASA's Star Trek RPG noted Subspace Radio at WF25 (which would be on the TOS/OCU scale).
The rec.arts.startrek.tech list's Warp Velocities FAQ (by Joshua Bell) lists WF9.9997 on the TNG/MCU scale, citing the TNG Tech Manual; that's ~198696 C, or roughly warp 58 on the old scale.