There's reason to be frustrated. You have a significant amount of time, effort and expense invested so far.
TVFool depends on accurate data from the FCC which includes transmitter power, antenna height above ground, antenna gain and antenna directional pattern. The algorithm used by TVFool also depends on terrain data from a third party source. If all the data is valid, the prediction still represents an estimate, because there are countless other variables that can influence real world reception conditions. Variable factors due to weather/atmospheric conditions, foliage, structures, real antenna system performance at a given frequency, etc. will play a roll in real world results, but are impossible to factor into a prediction.
When I look at a report entry that indicates a single digit NM value, a 2-edge path profile, a co-channel interference warning and the presence of an adjacent channel that's 15 dB stronger, I see plenty of reason to expect difficulty receiving that signal. If I decide that reliable reception of that signal is a must, I have to anticipate that I may need to go to 'extraordinary lengths' to achieve the desired result.
Though it may be obvious, it's worth pointing out to other readers of this thread, that just because a signal is at the top of a given report table, does not indicate that it is easy to receive... Occasionally we see reports with a list of signals, none of which are of sufficient strength or quality to expect reliable reception. In your specific case, all signals shown on your report are challenging. You're doing remarkably well with many signals that others with just a few trees nearby would find impossible to receive reliably.
As I recall, the 91XG and an Antennacraft Y10713 among the antennas suggestions made in your first thread. The Y10713 looks like it will soon disappear from the market place due to the Radio Shack bankruptcy. If I was in your area, I'd actually consider buying one or even two Y10713's. A pair of Y10713's correctly
ganged could offer a measurable improvement in gain over the CA8100 at real CH-12. The question of, 'would it be worth it?' is, of course, one that each person would have to answer for themselves.