Observed recombination frequency between two loci cannot exceed $50%$ in ordinary two-locus linkage analysis.

This does not mean that loci separated by large chromosome distances experience at most one crossover in half of meioses. The limit arises because we classify gametes only as parental or recombinant for the two endpoint loci.

Even and odd crossover histories can give different endpoint states

Start with a chromatid carrying

$$A;\cdots;B$$

and its homolog carrying

$$a;\cdots;b.$$

A crossover between $A$ and $B$ can switch the endpoint association, creating $Ab$ or $aB$ chromatids.

A second crossover in the same interval can switch the endpoint association again. Depending on which chromatids participate, multiple crossover events can therefore produce endpoint combinations that look parental even though recombination occurred within the interval.

The important lesson is:

$$\text{observed recombinant fraction}\neq\text{number of crossover events}.$$

Why widely separated loci approach 50%

As loci become farther apart, one or more crossovers between them become increasingly likely. But repeated crossover opportunities progressively erase information about the original phase.

At the limit, the four gamete types occur approximately equally:

$$AB,\quad Ab,\quad aB,\quad ab\approx\frac14\text{ each}.$$

Then parental gametes together make up about $1/2$ and recombinant gametes together make up about $1/2$:

$$r\rightarrow0.5.$$

A measured recombination frequency near $50%$ therefore means the two loci behave as if unlinked in a two-locus cross. They may truly be on different chromosomes, or they may be far apart on the same chromosome.

Pairwise recombination underestimates long map distances

Suppose an interval has a true cumulative genetic length greater than $50\ \mathrm{cM}$. Its endpoint recombination fraction can still never exceed $0.5$.

Therefore the simple conversion

$$100r\approx\text{cM}$$

cannot recover long distances directly from a single pairwise recombinant fraction.

This explains why geneticists map long chromosome regions using multiple intermediate markers. Short adjacent intervals can be estimated more faithfully and then added along the chromosome.

The 50% limit is thus an information limit of endpoint classification, not a physical upper bound on crossover activity.