It might be helpful to look at the paper by Martin Colloms on his web site. He is the author of HIGH PERFORMANCE LOUDSPEAKERS as well as being a speaker designer and audio critic. It is too long to reproduce but in part it states:
2.5 Anecdotal Consumer Press Reporting
I have considered that informal, anecdotal reports from reviewers concerning the advantage of addon
super tweeters to be rather suspect since I feel that generally the basis of their evaluations has
generally been so flawed that the results may be discounted. The difficulties encountered are
manifold and only a few need be considered to confirm my negative view. Firstly the test for
response extension benefit will only be valid if the extended response is achieved without affecting
the performance in the existing audible range.
Testing for a subtle effect, which may well be barely audible, is a manifest nonsense if it changes
the uniformity and loudness in the already operative treble range. Yet this is what is happening in
these tests. So far, no commercially available add-on tweeter and matching crossover can avoid
this fundamental error. Note that when such a driver and crossover is patched on to an existing
audio chain, as it often is, it will inevitably change the loading on both the cable and the amplifier,
and thus very likely impart another audible difference.
These super tweeters typically operate in parallel with the existing tweeter over about an octave
bandwidth, and may destructively interfere with the primary tweeter output. Thus there is a potential
to impair as well as alter the results. Sadly, some critics are so pleased to have heard a difference
they are tempted simply to judge it as an improvement. Often the crossover is a simple capacitor
feeding a metal dome tweeter, perhaps with a beryllium composite diaphragm. Such a crossover is
something of a disaster since a calculated 20kHz crossover, comprising a single capacitor for a
nominal 5ohm rated tweeter, provides the response shown in Figure 1. The intrinsic output is
compared with the crossover objective, which is seen to be markedly different from the practical
result generally obtained with such a single element filter. The cause is the complex impedance
presented by the high frequency driver compared with a plain resistor load. Even with more
complex, higher order filters, the practical crossover points for super-tweeters are often placed well
into the audio band. It is not surprising that audio professionals dismiss such published subjective
results, which often seem to be produced in support of media and equipment marketing.