I think they used paired data for their significance tests (at least they should have), but used the absolute values for the abstract. If they did use grouped data, then this could be a reason why differences between conditions were not significant.
Edited to add:
I just looked up the original article and it seems that they indeed worked on grouped data. Moreover, they applied a Bonferroni correction which is an overcorrection (although not everybody agrees with that but in this case I certainly believe it is - OT to expand on it). Considering that that a more appropriate data-analysis might show true differences between the cadence conditions, the study does not really convince me at this point.
Edited to add:
I just looked up the original article and it seems that they indeed worked on grouped data. Moreover, they applied a Bonferroni correction which is an overcorrection (although not everybody agrees with that but in this case I certainly believe it is - OT to expand on it). Considering that that a more appropriate data-analysis might show true differences between the cadence conditions, the study does not really convince me at this point.