Here's an answer for your question 1. which has two parts:
a) [An example of a] Banach *-algebra that is not a C*-algebra for which there exists a positive linear functional (it takes $x^*x$ to numbers $ \geq 0$) that is not norm continuous.
As far as I know all “explicit” counterexamples are based on Dixon's construction, and they are quite complicated. In 1982 H.G. Dales proved:
There is a Banach *-algebra with a norm-discontinuous trace, i.e. a positive linear functional $\tau$ such that $\tau(ab) = \tau(ba)$ for all $a,b \in A$.
See Example 4 of H.G. Dales, The Continuity of Traces, in Radical Banach algebras and automatic continuity, Lecture Notes in Mathematics, 975, Springer Verlag, (1982), 451–458. To understand the construction in that article you will need a copy of P.G. Dixon's paper, Non-separable Banach algebras whose squares are pathological, J. Functional Analysis, 26 (1977), 190–200.
Remark. That this example must be rather complicated is not really surprising since it is impossible to “write down” a discontinuous linear map, and it is obviously even harder to write down a discontinuous linear map satisfying further algebraic constraints: a discontinuous linear map necessarily involves the axiom of choice in a strong form, e.g. existence of algebraic bases, since it is consistent with ZF+DC that all linear maps between Banach spaces are continuous. In fact, the (non)existence of discontinuous homomorphisms on $A = C(K)$ is independent of the usual axioms of set theory, see also Andres Caicedo's answer here.
b) Apparently if a Banach *-algebra so much as even has a bounded approximate identity, then all positive linear functionals are continuous. Does anybody have a proof of this?
Yes. More generally, Theorem 5.5.10 on page 698 of H.G. Dales, Banach Algebras and Automatic Continuity states the automatic continuity of a positive linear functional on a Banach algebra with a not necessarily continuous involution and a bounded approximate left identity. Here's an outline of the argument in the simpler case of an isometric involution:
Suppose that $A$ is a Banach *-algebra with a bounded approximate identity.
Arguably the most important single result in automatic continuity theory is the Cohen-Hewitt factorization theorem. In one of its most basic forms it can be phrased as follows:
Let $A$ be a Banach algebra with a bounded approximate left identity $(u_i)_{i \in I}$ and let $M$ be a left Banach $A$-module. For $m \in M$ the following are equivalent:
There are $a \in A$ and $n \in M$ such that $m = an$
$\lim_{i} \lVert u_i m - m\rVert = 0$.
A short proof (very close to Hewitt's argument) can be found in Cigler–Losert–Michor, Banach modules and functors on categories of Banach spaces, Theorem III.1.15 on page 108 (they are assuming that the approximate identity is in the unit ball of $A$ but the proof is easy to adapt to the more general statement given here).
Over time the factorization theorem has been sharpened in many ways and entire books were written about it: R.S. Doran, J. Wichmann Approximate Identities and Factorization in Banach Modules. See also Chapter 2 of Dales's comprehensive Banach Algebras and Automatic Continuity.
Some further remarks:
If all $m \in M$ satisfy the conditions of the factorization theorem then $M$ is called an essential $A$-module.
Since the definition of an approximate identity is property 2. for all $a \in A$ when $A$ is considered as a left $A$-module, one immediate consequence of the theorem is that every $a \in A$ factors as $a = bc$. For $A = L^1(G)$ this is Cohen's original form of the factorization theorem.
An easy lemma on essential modules $M$ is that the space $c_0(M) = c_0(\mathbb{N},M)$ of null sequences in $M$ with the norm $\lVert (m_n)_{n \in \mathbb{N}}\rVert_{c_0(M)} = \sup_{n \in \mathbb{N}} \lVert m_n \rVert_M$ is an essential module (check that property 2. is satisfied for $c_0(M)$). This implies that every null sequence $m_n \to 0$ in $M$ can be factored as $m_n = a x_n$ with $a \in A$ and $x_n \to 0$ in $M$.
[Incidentally, this can be used to show that a right $A$-linear map $\varphi\colon A \to N$ into a right Banach $A$-module is automatically continuous by observing that $a_n \to 0$ implies $a_n = ab_n$ with $b_n \to 0$ so that $\varphi(a_n) = \varphi(a)b_n \to 0$.]
The proof of continuity of positive linear functionals on a Banach *-algebra with a bounded approximate identity is now relatively easy:
Recall the Cauchy-Schwarz inequality for positive linear functionals $f$:
$$
\lvert f(ab a^\ast)\rvert \leq f(aa^\ast) \cdot \sqrt{\rho(bb^\ast)},
$$
for all $a, b \in A$, where $\rho(x) = \lim_{n\to\infty} \lVert x^n\rVert^{1/n}$ is the spectral radius of $x \in A$.
It follows from this that $x \mapsto f(a x a^\ast)$ is continuous for all $a \in A$ and thus $x \mapsto f(ax b^\ast)$ is continuous for all $a,b \in A$.
Now let $x_n \to 0$. We want to show that $f(x_n) \to 0$. By the factorization theorem (remark 3. above), we can write $x_n = a c_n$ with $c_n \to 0$. Now $c_n^\ast \to 0$, so we can write $c_n^{\ast} = b d_n$ with $d_n \to 0$. Thus, $x_n = a d_n^\ast b^\ast$ with $d_n \to 0$, so $f(x_n) = f(a d_n^\ast b^\ast) \to 0$, as desired.
What follows (and all references given) will be based on Chapter 16 of the book "Introduction to Functional Analysis" by Meise & Vogt (1997). There you can also read up on the key spaces in this matter if you are not all too familiar with them yet: the compact operators $\mathcal K(\mathcal H)$ and the trace class $\mathcal B^1(\mathcal H)$ (in the above book denoted by $S_1(\mathcal H)$ for the Schatten-1-class). The inclusions between these spaces in infinite-dimensions read $\mathcal B^1(\mathcal H)\subsetneq\mathcal K(\mathcal H)\subsetneq\mathcal B(\mathcal H)$ (in finite-dimensions they all coincide).
One needs the trace class, obviously, for the trace $\operatorname{tr}(\omega B)$, $B\in\mathcal B(\mathcal H)$ to make sense beyond finite dimensions. One can actually show that every continuous linear functional $\tau:\mathcal B^1(\mathcal H)\to\mathbb C$ (i.e. every dual space element $\tau\in(\mathcal B^1(\mathcal H))'$) is precisely of the form
$$
\tau(A)=\operatorname{tr}(AB)\qquad\text{ for some }B\in\mathcal B(\mathcal H)\text{ and all }A\in\mathcal B^1(\mathcal H)\,,\tag{1}
$$
cf. Proposition 16.26; for short $(\mathcal B^1(\mathcal H))'\cong\mathcal B(\mathcal H)$. Also the trace class in infinite dimensions is not reflexive (Corollary 16.27) meaning that $\mathcal B^1(\mathcal H)\not\cong (\mathcal B^1(\mathcal H))''\cong (\mathcal B(\mathcal H))'$. In other words, not every dual space element of $\mathcal B(\mathcal H)$ is of this trace form (1).
Instead one can show that every $\varphi\in(\mathcal K(\mathcal H))'$ can be written as $\varphi(K)=\operatorname{tr}(KA)$ for some $A\in\mathcal B^1(\mathcal H)$ (Proposition 16.24). So if one restricts oneself to the compact operators, every functional is of trace form again--but $\mathcal B(\mathcal H)$ is obviously "way larger" than $\mathcal K(\mathcal H)$.
To conclude, all I said can then be nicely summarized in one identity:
$$
\boxed{(\mathcal B(\mathcal H))'\supsetneq(\mathcal K(\mathcal H))'\cong \mathcal B^1(\mathcal H)}
$$
which--reformulated--says that there exist "non-normal states" in infinite dimensions (those are precisely the states which are not weak-* continuous).
Best Answer
Nothing. I cannot comment on the specific text you are reading since you don't say what it is. But pure states are irrelevant towards constructing the universal representation.
What you need to do to construct the universal represention is the following:
first you show that states separate points. More specifically, given $a\in A$ there exists a state $\varphi$ such that $|\varphi(a)|=\|a\|$. This is achieved using the Gelfand representation.
Then you construct the GNS representation for a single state.
Then you construct the direct sum of all (unitary classes of) GNS representations coming from a state. Because the states separate points, this representation is faithful, and you are done.
There is not need for pure states in the above.