Well, it is. And the reason is that the initial intervention is assumed to be uncaused - hence undetermined - while everything after that point is assumed to be determined (albeit in a non-linear fashion).
So, the flapping of a butterfly's wings is implicitly supposed to be uncaused - by the free will and agency of the butterfly; but from the moment those wings have flapped there is assumed to be no further agency, but events unroll inevitably (e.g. toward a hurricane somewhere else...).
This bizarre assumption probably comes from the origins of the Butterfly Effect in computer programming - including analogue computing: the programmer (implicitly assumed to be an undetermined agent) stands outside the system he is manipulating, but his set-up and interventions unroll deterministically.
In the real world, however, the validity of the Butterfly Effect depends on agency. In the first place, the Butterfly wing-flap may not be an act of agency, but may itself be a determined consequence of prior causes - in which case, the Butterfly wings do not cause anything at all - they are merely one link in a multi-factorial causal chain of unknown origin.
On the other hand, if the Butterfly Wing-flap was due to Butterfly agency, hence uncaused; then we have acknowledged the reality of agency in the world, and cannot afterwards coherently assume that agency is necessarily absent from the system from that point (i.e. we cannot assume that the entirely of the system is deterministic if the programmer is not).
(In computer programming terms, if the initial programmer has the free will/ agency needed to set-up a deterministic system and set it into motion; then other programmers may be a part of the system and divert it unpredictably (because interventions were uncaused) - or further programmers may intervene from outside the system to make changes to the outcome.)
In a softer version; the Butterfly Effect is mistakenly used to refer to the ancient observation that small causes may have large outcomes, and large outcomes may depend on small causes; as in the old nursery rhyme 'For want of a nail' - which traces the death of King Richard III in the Battle of Bosworth to the lack of a single horseshoe nail.
But even here, the Butterfly Effect gets misused and generally misleads. What is valid about the Butterfly Effect is to explain one reason why empirical and statistical methods that work well over the short term become worthless, rapidly as the timescale is extended. The longer-term is therefore either unpredictable; or else requires the measurement of entirely different variables.
For example weather forecasting, in the UK, sometimes works well over a few hours or a day, based on measures such as wind speed and direction and satellite pictures (and predictive programming); but becomes useless over a few days, and is worthless for medium- and long-term forecasting. Medium/ Long term forecasting either can't be done at all, or seems to be done better by focusing on different measures; for example the activities and doings of the sun and moon, somewhat as Piers Corbyn does. And the worthlessness of standard weather forecasting methods for medium/ long term forecasting based-on wind etc is utterly unaffected by the use of ever bigger superexpensive supercomputers (which are periodically successfully demanded by the state funded forecasting services, and which never work because they never can work).
On the whole, however, it is clear that the Butterfly Effect has approximately zero real-world relevance or validity - whether defined strictly, or as the term is used in pop culture.
