Consider a single finite length wire, bent in an L shape, 1/4 wavelength of some given frequency F0 from the end. A current impulse, say driven by a single spark from a spark transmitter, will indeed be seen traveling in the reverse direction at the corner of the L at time roughly half the period of F0 later. However this "bounced" current impulse will be reduced, due to a portion "lost" due to wire resistance plus EM energy loss due to coupling into free space (or any nearby or distant conductors). This loss is usually less than 100% for typical wires of typical lengths in free space.

Now add another L shaped conductor nearby in a mirrored arrangement. Some of the "bounced" current might flow as displacement current across the two L corners to the other 1/4 wavelength L stub, due to the mutual inductive field.

So some of the current (the portion not "lost") indeed flows back down the feedline, which, summed with current send up the feed line, adds up to a current to voltage ratio, which we can call the resistive impedance of the feed point.

Now consider a sinusoidal drive of frequency F0 to be the sum of a bunch of these impulses, modulated in amplitude by sin(F0).

Consider a single finite length wire, bent in an L shape, 1/4 wavelength of some given frequency F0 from the end. A current impulse, say driven by a single spark from a spark transmitter, will indeed be seen traveling in the reverse direction at the corner of the L at time roughly half the period of F0 later. However this "bounced" current impulse will be reduced, due to a portion "lost" due to wire resistance plus EM energy loss due to coupling into free space (or any nearby or distant conductors). This loss is usually less than 100% for typical wires of typical lengths in free space.

Now add another L shaped conductor nearby in a mirrored arrangement. A portion of the "bounced" current might flow as displacement current across the two L corners to the other 1/4 wavelength L stub, due to the mutual inductive field.

So some of the current (the portion not "lost" and not sent a displacement current) indeed flows back down the feedline, which, summed with any current send up the feed line, adds up to a total current to voltage ratio, which we can call the impedance of the feed point.

Now consider a sinusoidal drive of frequency F0 to be the sum of a bunch of these impulses (infinite in number, infinitesimal in size), modulated in amplitude by sin(F0).