I agree this seem contradictory but what i concluded from these sections is:
-Bottom-up is based on management forecasts which tend to me more optimistic (hence even heading into recession, management are likely to be optimistic).
Top-down is based on econometrics which work well as long as the estabilished relationships are stable- y= mx +c. When there is a sudden change in trends/relationships, econometrics wont pick that. Hence slow to such changes
for a top down model - you are using historical data - and that does not have enough time to change and reflect the current trends. so your econometric model is based on dated data - so it would be slow to detect changes that are ongoing.