Consider the following:
The Grothendieck topology is not a topology.
A Riemannian metric is not a metric.
The notation $\lim_{\to}$ is a co-limit, not a limit.
(Two maybe different examples)
The complex plane is (for algebraic geometers) a curve. Even worse, the Riemann sphere is also a curve.
An affine group scheme is a representable functor.
What are some mathematical objects that follow this unfortunate pattern?
On
A manifold with boundary is not necessarily a manifold (and when it is, it's arguably not "with boundary").
On
Elliptic curves are not related to ellipses, but to elliptic integrals.
Additionally, the monster group is not a monster, it's just misunderstood.
On
I'm not sure whether this qualifies, since "series" does not have another meaning in group theory so far as I'm aware, but it is annoying given the potential for confusion between sequences and series in elementary analysis.
Let $G$ be a group. A (subgroup) series for $G$ is actually a sequence of subgroups $H_{1},H_{2},\ldots,H_{n}$ of $G$ satisfying $$1<H_{1}<H_{2}<\cdots<H_{n}=G.$$
Of course, this extends to any particular kind of series, e.g., normal series, composition series and chief series.
On
The Peano-Jordan measure isn't a measure (the measurable sets don't form a $\sigma$-algebra), because of this reason some authors prefer the term "jordan content"
On
From MacPherson's manuscript Intersection homology and perverse sheaves:
The first thing to know about perverse sheaves is that they are neither sheaves nor are they perverse.
Goresky has explained at MO the etymology of the term, but it still counts in my opinion as a bit of a terminological stumbling block for the unwary. (Although a certain amount of wariness is called for around anything explicitly called "perverse," so ... touche I guess?)
From Wikipedia: