We study the integrality gap of the natural linear programming relaxation for the Bounded Color Matching (BCM) problem. We provide several families of instances and establish lower bounds on their integrality gaps and we study how the Sherali-Adams "lift-and-project" technique behaves on these instances. We complement these results by showing that if we exclude certain simple sub-structures from our input graphs, then the integrality gap of the natural linear formulation strictly improves. To prove this, we adapt for our purposes the results of Füredi [Combinatorica, 1(2):155-162, 1981]. We further leverage this to show upper bounds on the performance of the Sherali-Adams hierarchy when applied to the natural LP relaxation of the BCM problem. say that edge e "has" color C j if e ∈ E j . Let C = ∪ i=1,...,k C i be the collection of all color classes. Each color class C j is associated with a positive number w j ≥ 1. Our goal is to find a maximum (weighted) matching M that contains at most w j edges of color C j i.e., a matching M such that |M ∩ E j | ≤ w j , ∀C j ∈ C.In [46] an LP-based approximation algorithm with approximation ratio 1/2 was given for the BCM problem, which matches the integrality gap of the natural LP relaxation for this problem. The algorithm is based on the elegant technique by Parekh [40] which gives an inductive process to write any basic feasible solution of the relaxed LP as an approximate sparse convex combination of integral solutions. The result holds for any bounds w j ≥ 1, integral or otherwise, since the analysis does not make use of the fact that w i ∈ Z + , ∀i, only the fact that w i ≥ 1 (otherwise the integrality gap could be unbounded). It has been further generalized by Parekh and Pritchard [41] to uniform hypergraphs.A very natural question occurs: a negative result based on a bad integrality gap instance rules out the possibility of a good relaxation-based approximation algorithm. But this holds only for the particular relaxation that we use. What about other, more complicated and sophisticated relaxations? As an illustrative example, if we take the normal (degree-constrained) relaxation for the classical matching problem, which has integrality gap of 3 /2, and enhance it with the blossom inequalities, we get an exact formulation of the convex hull of all integer points for the matching problem [16].Given the apparent difficulty of identifying stronger/tighter linear relaxations for combinatorial optimization problems, a large body of work has been dedicated in recent years to identifying systematic techniques to enhance the quality of a given linear (or semi-definite) program with valid inequalities (inequalities that are satisfied by all integral points). The hope is that the part of the polyhedron responsible for the bad integrality gap example will be eliminated. Many such "lift and project" methods have been proposed so far, in particular by Sherali and Adams (SA) [45], by Lovász and Schrijver (LS) [29], by Balas, Ceria and Cornuéjols (BCC) [5], by Lasserre [25] and by B...