Another retired engineer type here. I use Quickload (QL) quite a bit. I have checked it against published load data a lot.
The tool can be very good for most applications if the user knows what he is doing. The biggest exception for QL accuracy is probably revolvers. QL does not even try to address cylinder throat length and cylinder gap. The next item is that QL will never be able to do is predict "spikes" from non-uniform ignition/burn. I also do not trust QL with high powder compression and there is at least one powder (Lil Gun) that just does not seem to perform as modelled.
Excluding the above, if you do real good at the inputs, my bet is that you get better results with QL than with using published data that is not quite exact for your load.
The most significant "avoidable errors" will usually be related to percent fill. Case capacity, case length, bullet length and COAL are the raw inputs to get this one parameter. If you get percent fill very close to either QL or to the as tested configuration for published load data you get rid of a lot of potential error.
The remaining factors are things like chamber dimensions (loose vs tight, gas leakage around the bullet before engraving, and distance for bullet movement before engraving), bullet resistance as it engraves in the rifling, primer affects, and powder burn rate variations.
With QL, you can "benchmark" your load by measuring velocity, adjusting it to muzzle velocity, and then going back in and adjusting the QL value for powder burn rate (Ba). This will tend to do a "one size fits all" correction for any inputs that were not quite right.
With good inputs and a good benchmark, the QL estimate of pressure for most loads will probably have an error that is in the same ball park as the error with something like what Larry uses. I say this not because the QL predictions will be that "great". I say it because of the accuracy limitations with the available measuring system for something as messy as chamber pressure.