So you sometimes don't use the absolute tops of swing highs or the bottoms of swing lows? And you look for key markers such as wide range breakout bars and gaps too?
Most critics will not bother to do the work but David Halsey's Trading the Measured Move is a good starting point to understand the why and where. Traders, including myself are likely to make adjustment to Halsey's trade play with respect to time frames, zones and trade management according to our own testing and experiences. As I've said before (to no lack of flack), I use a looser zone of 1/3 to 2/3 of the defining leg for retracements and look for price action in that zone to trigger a trade. For targets, fib extension grids off the defining leg are effective. Don't believe it?...watch what happens at fib levels. Either way suit yourself.
You are going on and on about drawing them correctly and attributing false conclusions about their effectiveness by not being done so. So how do you draw them correctly?
The paper has The the high and low points in the past 100 trading days are considered the zero and one levels and Fibonacci ratios like 0.382 and 0.618 of the differences around the found zero and one levels are considered for retracements and projections. As new high and low points are found, the price levels for Fibonacci ratios change. Then, conventional-type entries like 50EMA ≥ 100EMA and a few bars after a pivot low are used with the Fibonacci retracements and projection levels used for stops and targets. The rules don't do well in backtesting. I don't think the method of finding the zero and one points and applying Fibonacci ratios based on their difference is much in the "spirit" of Fibonacci because the zero and one points aren't necessarily near price wave peaks and troughs.