Correct.   I have a PIC that reads timing data from a GPS and at the right timing uses a set of pre-stored WSPR symbols to calculate the frequencies for the tones and sends the resulting data to an AD9852 DDS.  Different symbol sets (for the two messages) and frequency data is stored for each of the two modes.

A pseudo random (PN) sequence that repeats every 30 with enough '1's placed in it to give the desired duty cycle is used to decide the Tx pattern, to decide if a transmission of either type will occur in a particular slot.  The same PN sequence is used for -2 and -15 modes, although different repeat squences and different lengths could just have easily been implemented for each mode.

At every two minute slot as read from the GPS timing data, and assuming a -15 is not currently underway,  the position in the pattern is examined to see if a -2 transmission should start.   Similarly, at every 15 minute slot if a -2 transmission is not underway, the position in the -15 pattern decides if that starts.

Although both -2 and -15 transmissions use the same PN sequence, they are stepped though based on the results from stepping theough the other and a semi-chaotic combination results giving a pretty random combination of -2 and -15 transmissions with the correct duty cycle on each.    The only contention is at :00 and :30 minutes if both are scheduled to Tx, in which case  -15 mode dominates.