indeed if the air pressure is 29.957 the actual pressure might be 29.927. but then it will not show 29.987 with the next measurement or any time soon (when actual air pressure doesn’t change) because the relative accuracy is only 0.003". It might fluctuate by that little amount.
If the real air pressure rises 0.003 you should notice it. If the air pressure for example raises every 30 minutes by 0.00333, you would see a sequence like
29.957 29.960 29.963 29.967 29.970 29.973 … all very accurate relative to your first measurement of 29.957. They might all be off by 0.03", but the relative difference is accurate and you might conclude that pressure is on the rise.
If these values are rounded to two decimals it looks like
29.96 29.96 29.96 29.97 29.97 29.97 which cannot lead to any accurate conclusion about the trend as for these measurements the real air pressure might have been
29.963 29.965 29.965 29.967 29.966 29.966 and you would have gotten the same rounded numbers and the difference is just noise.
of course the first sequence looks nicer in hPa or mBar.
edit: in addition if you would calibrate the device at some point in time to the exact value (which can easily be done after plenty of measurement that show for example that the pressure is always 0.03 too high), you will have a very accurate system that shows you the pressure with an absolute precision of 0.003. It might drift over time but re-calibrating often enough to compensate for a possible drift would solve that. That is exactly what weatherflow is doing.